Featured Video Play Icon

Socialized Attention and Situated Agency

Bryce Huebner (Georgetown University)

[Jump to comments]


Across numerous sub-fields in philosophy, people are beginning to address the impact of evaluative expectations on what we know, and how we experience the world. Philosophers working on pragmatic encroachment, epistemic risk, and cognitive permeation, have all highlighted the social contingency and normativity of human thought. But there’s a great deal of dispute over how evaluative expectations affect cognitive processing, if they actually do. My goal in this paper is to explore a perspective that places evaluative cognition, social scaffolding, and ongoing risk assessment at the center of human cognition, but does so in a way that doesn’t presuppose the more exciting claims of philosophers who defend synchronic forms of cognitive permeation. I build on my earlier argument for the claim that our evaluative attitudes gradually attune to local environmental regularities (Huebner 2015; 2016); but here, I expand on the variety of important (though not exhaustive) roles that culturally learned expectations can play in shaping attentional strategies, producing culturally situated affordances, and constraining judgments about what we should do and how we should act.

The claims that I make in this paper are intended to be empirically assessable, even where I rely on anecdotal claims and theoretical assumptions. So I acknowledge that my claims in this paper might be falsified, and that other approaches may prove more promising in the long run. So be it! My aim is to get clear on the nature of socially-situated agency, not to offer the final word on issues of the socially situated nature of thought and behavior. But I hope that this can be the beginning of a discussion, which can lead us to better ways of linking the insights that can be gained by looking-down to neuroscience, and looking out to sociology and cultural psychology.

1. Racializing perception?

Over the past decade (maybe a bit less), philosophers have become increasingly interested in the claim that the brain is a hierarchically organized predictive machine, which mines past experience for information that will allow it to anticipate its future states (see Clark 2015, Frith 2007, and Hohwy 2013 for reviews). One part of this claim is commonly accepted across much of contemporary cognitive science: The human brain is a hierarchically organized system; mechanisms closer to the sensory periphery process more precise and concrete sources of information; and processing becomes more abstract and conceptual as information is propagated upward through this hierarchy. Proponents of the predictive brain hypothesis, defend a more controversial set of claims about this hierarchically organized system. They hold that at each level of processing, systems attempt to predict the inputs that they will receive, and each system generates error-signals when surprising data are encountered; as these error-signals flow upward through the computational hierarchy, they recruit predictions that will better accommodate the incoming data; and as predictions flow downward through the computational hierarchy, ambiguities in the incoming data are resolved by imposing expectations about what will happen next. Over time, this bi-directional flow of information allows the brain to search for “the linked set of hypotheses that, given the current raw sensory data, make that data most probable” (Clark forthcoming).

The most familiar version of this hypothesis also suggests that expectations permeate the structure of experience at every stage of processing. What we experience is always filtered through our brain’s expectations about what we are likely to experience; and while the world is saturated with useful information, which anchors us to the world, these expectations impose a great deal of structure where the incoming data are ambiguous or underspecified (which is pretty often, in a constantly changing world). One way to make this claim exciting is by pointing to cases where expectations shift our immediate experience of the world. And there are plenty of cases where this seems to occur. For example, it’s difficult to parse sinewave speech, but sinusoidal variants of English language sentences are intelligible if you’ve heard the transformed sentence in English, and know what to expect. More intriguingly, knowing that a sentence is the beginning of a fairytale is sufficient for many listeners to impose a stored set of expectations on perceptual data, resolving ambiguities in the signal on an initial listen (see Clark 2015, 54-55; and for a demonstration, see http://goo.gl/Em2zgp).

Expectations also seem to impact perception in ways that cross into socially salient aspects of our lives (and this is the primary reason why I’m interested in questions about expectation-driven processing). As many philosophers have noted, Daniel Levin and Mahzarin Banaji (2006) found that expectations regarding skin color seem to influence perceptual experiences, leading people to experience a phenotypically Black face as darker than a phenotypically White face, even if they are matched for overall luminance. In a sense, these data aren’t surprising, as people routinely report seeing Black children as adults, and seeing Black adult as dangerous assailants (Goff et al 2014). In light of this fact, we might assume that race shapes what people see. And Lisa Feldman Barrett (2015) has recently suggested, this may be the reason why a barely glimpsed object might look like a gun if you’re already expecting things to go badly. She claims that affective states shape ongoing patterns of action-oriented cognition, as well as more accessible trains of thought; and she claims that when these states are bundled together, this can generate distorted forms of perceptual experience.

Data for this hypothesis seem to derive from the well known research into the role of racial expectations in shooter bias tasks, which force participants to quickly decide whether to shoot a person or not, depending on whether they are holding a gun or an innocuous object (e.g., a cell phone or a wallet). However, in a meta-analysis of 42 experiments on shooter bias, Yara Mekawi and Konrad Bresin (2015, 124) found that:

Compared with White targets, participants were quicker to shoot armed Black targets, slower to not shoot unarmed Black targets, and were more likely to have a liberal shooting threshold for Black targets. They were not, however, more likely to be sensitive to or have a higher false alarm rate for Black (versus White) targets.

But these patterns are affected by some cultural factors. Specifically, people who live in states with more permissive gun laws have lower shooting thresholds, and they are more likely to mistakenly choose to shoot; and people who live in cities with lower proportions of White inhabitants display larger biases against Blacks. Together, these data support the claim that racially structured expectations can facilitate or inhibit people’s responses in shooter bias tasks, but not in the way that is typically assumed. People are more likely to respond by choosing to shoot a Black person—period. And this should worry all of us. But the fact that there are no significant differences in false-alarm rates suggests that people are not seeing innocuous objects as guns when they are in the hands of Black people, even though they are more willing to choose to shoot them. Finally, higher rates of gun ownership make people trigger happy and inaccurate; and people living in cities with larger non-white populations have lower shooting thresholds and make more mistakes when the person they see is Black (Mekawi & Bresin 2015). Things are bad! But not in a way that provides support for the most common versions of the hypothesis that expectations permeate experience.

A recent study by Joshua Correll and his colleagues (2015) provides a bit more evidence about the role of expectations in these kinds of shooter bias tasks. Using an eye-tracker and a standard shooter bias task, they found a significant effect of race on the visual angle between the object to be categorized (the weapon or the innocuous object) and the fovea at the time of response. More specifically, they found a significantly higher visual angle for Black-with-weapon conditions as compared to White-with-weapon conditions. This suggest that people need more visual information to judge that a White person is holding a gun than they needed to judge that a Black person is holding a gun; and they may be more likely to choose to shoot a Black person because they assume that Black people are more likely to be dangerous. To my mind, the most plausible reading of these data is that participants need less visual information to decide in favor of the hypothesis that a particular Black person is dangerous, because they tend to assume that Black people are more dangerous, simpliciter. But if this is correct, it is just be evidence that the amount of information we need to make a decision depends on how sure we are in the first place, and perhaps how committed we are to getting things right (cf., Machery 2016, 64).[1]

Likewise the beautifully constructed experiment by Levin and Banaji (2006) cannot demonstrate that expectations permeate experience, leading us to see things differently; the problem is that “the luminance is not similarly distributed in the bottom half of the two faces”, and it’s possible that people have different experiences in this case because they attend to features of the faces that are differently illuminated (Machery 2016, 68-69). If this is right, then expectations are structuring where people look, and yielding the predictable downstream effects on subsequent processing. Here too, an attentional reading of the data looks more plausible. Indeed, cases of cognitive permeation are really hard to find, even outside of these politically charged domains. Many studies purporting to show that expectations affect perception have failed to replicate; many others reveal more general effects on judgments, or more specific effects within a single modality (Gross et al 2014); and most existing studies fail to rule out effects of attention, which shape the inputs to perceptual processing but do not change perceptual processing (Firestone & Scholl 2016).

Still, I think that there’s something right about the form of affective realism that Barrett (2015) defends. Affective expectations can increase our awareness of potential threats, they can generate negatively valenced states, they can enhance the accessibility of associations between race and violence, and they can trigger the construction of forward-looking plans for addressing potential threats (Wilson-Mendenhall et al 2011). When all of these effects are bundled together, they yield racialized behavioral and psychological dispositions, but not necessarily because there is a shift in what we experience; in many cases, this is simply because of a shift in what we pay attention to. They also modulate the amount of information we need to support our hunches. To understand this effect, I think we need to look more closely at the relationship between culture, affective valence, and attention. And in the remainder of this paper, I argue that experiencing particular kinds of socially structured environments can yield shifted patterns of person-level attention.

2. Socializing and shifting attention

In a series of recent papers, Shinobu Kitayama and his colleagues have argued that attentional biases arise through a process of cultural evolution and cultural attunement. And I think that they are right that active, repeated, and ongoing participation in culturally scripted behavior will lead to the gradual development of “attention allocation strategies that are consistent with local cultural assumptions” (Park & Kitayama 2011, 77). On this perspective, cognition is always situated, as what we attend to, what we think about, and what we will remember depends on facts about the social environments that we are chronically immersed in. This doesn’t require expectations to permeate experience. But it does require diachronic shifts of attention, which modulate the salience of particular perceptual phenomena. And it requires a sort of neural plasticity that can explain social attunement to local environmental regularities (Kitayama, Park, and Cho 2010).

According to the socialized attention hypothesis, people are “socioculturally shaped shapers of their environments” (Markus & Kitayama 2010, 421). As social niche constructors, our understandings of what the world affords are shaped by the social and material structures that we simultaneously create and inhabit. People are rewarded for forms of social engagement that accord with local norms, and they are criticized for acting in ways that are socially deviant. As a result, people learn to categorize information in ways that are culturally sanctioned; and since culturally deviant forms of categorization are rarely reinforced, many psychological processes will tend to reflect the worlds, contexts, and social systems that people are chronically immersed in (cf., Dasgupta 2014). People also tend to imitate the actions of successful or prestigious people, in hopes of uncovering successful strategies that will help them to navigate their local environments; and as a result, people tend to develop locally useful strategies for thinking about, navigating, and evaluating the constraints imposed by the material structure of the world that they inhabit. In this way, mechanisms of cultural learning propagate values and behavioral patterns through social groups (cf., Richerson & Boyd 2008; Henrich & McElreath 2003; Norenzayan et al 2016; Slingerland 2014); and as these patterns stabilize as successful behavioral and cognitive strategies, they impact the material structures and ideological practices that surround us, and shape the attitudes and attentional strategies that we acquire (Kitayama & Uskul 2011, 422).

Some of the most salient regularities that cultural psychologists have uncovered are those that continually emerge in predominantly middle-class, predominantly White, North Americans (most of whom are college students; cf., Henrich, Heine, & Norenzayan 2010 on this WEIRD population). For many people in this demographic, everyday interactions foster voluntaristic and independent accounts of agency, And numerous studies have revealed that White, middle class, North Americans tend to converge on attentional strategies that insulate their thinking from contextual factors, and focus their attention on discrete entities (Adams et al 2010; Markus & Kitayama 2010; Park & Kitayama 2011). This is not an innate disposition, and it’s not a fact about every middle-class White American. But in general, the environmental contingencies that shape experiences within this demographic are organized around ideas that highlight the importance of individual achievements, and that reinforce these ideas through everyday practices of praising and blaming individuals for their actions; so this culture sees value primarily in individual success, and provides ideological support for the assumption that self-interested actions will often yield such success (Markus & Kitayama 2010, 428).

More importantly, the focus on individual agency is scaffolded by ongoing engagements with “such concrete realities as mobility-affording transportation and communication infrastructure, the practice of ‘leaving home’ in young adulthood, the daily practice of eating from individual place settings, and residence in self- contained apartment units” (Adams et al 2010, 283). By routinely experiencing these social institutions and social opportunities, people develop preferences for “exploration, expression, and indulgence of unique, individual feelings”as one of the primary goods to be pursued (Adams et al 2010, 284). Consequently, they tend to “express a desire for mastery, control, achievement, choice, self-expression, or uniqueness” (Markus & Kitayama 2010, 421). And as a result, engagements with this material environment become entrenched as psychological and behavioral dispositions. Finally, as this desire becomes part of the subjective understanding of agency, it fosters schematic patterns for planning future activity.

Converging claims about social class have been advanced by Michael Kraus and his colleagues (2012). They propose that economic and social constraints impact what an individual knows, shaping everything from action tendencies to affective responses, and impacting the way that people think, feel, and relate to one another (Kraus et al 2012, 550). In line with a great deal of research in feminist philosophy, they argue that the experience of economic and social freedom leads to the development of cognitive strategies that are focused on internal states, and that treat these states as the dominant influence on thought and behavior. When people are chronically immersed in environments of relative abundance and elevated social rank, they “are free to pursue the goals and interests they choose for themselves”, and they can “pursue these goals and interests relatively free of concerns about their material costs” (Kraus et al 2012, 550). And as a result, middle-class and upper-class individuals tend to prioritize individualized selves, and assume that behavior is generally caused by individuals, instead of depending on contextual or situational factors.

By contrast, contexts where the stability of necessary resources is uncertain and unpredictable lead to the attentional prioritization of context and situations over individual states. People who live in less stable neighborhoods, who face constant economic instability, and who depend on constantly fluctuating institutional resources often experience the world as socially structured, institutionally constrained, and more limited in social opportunities (Kraus et al 2012, 549). As a result, people who are chronically immersed in these sorts of environments tend to develop attentional strategies that are sensitive to cases of overt social control, and they tend to be aware of the continual re-coding of their actions in accordance with the prefered frameworks of people in positions of social power (Again, echoing mountains of research from feminist approaches to standpoint epistemology).

These processes of cultural learning make it possible for people to acquire the attentional, evaluative, and cognitive skills that they need to succeed in their local environments. And in general, even though some kinds of attention and evaluation can be problematic, most people tend to converge on psychological and behavioral dispositions that are well-suited to their local environment. But the pervasive capacity for the cultural attunement of cognition is both a blessing and a curse. Once material and ideological structures are in place, individuals can exploit their attentional biases to minimize the cognitive effort that’s required to act in a socially accepted way. But just as often, patterns of exclusion and oppression can become entrenched in the material and ideological structures of our cities and social spaces. To explore this suggestion, I want to turn to some case studies, which highlight the ways that spaces, places, and situations can shape culturally attuned patterns of racialized thought.

3. White spaces and White Places

Consider the design of cities, and the organization of social space in the United States.[2] Most White people are exposed primarily to material and social environments that reflect their own interests and perspectives, and not to the interests and perspectives of people of color (DiAngelo 2011, 58; Moore 2008). This shouldn’t be surprising, and it’s not an accident. Practices of redlining by the the Federal Housing Administration starting in the mid-1930’s prevented people from getting home loans in the neighborhoods where Black people lived; and the mortgage industry as a whole quickly adopted similar lending practices, driving Black neighborhoods into poverty, preventing the establishment of racially mixed neighborhoods, and preserving the racial homogeneity of White neighborhoods (cf., Coates 2014; Madrigal 2014). In some respects, these practices have been scaled back (or at least gone further underground), but forms of bias in lending persist; and the recent housing crisis disproportionately affected recently integrated neighborhoods, as well as predominantly Black and Latinx neighborhoods, contributing to the re-entrenchment of de facto segregation (Badger 2016; Hall et al 2015). The upshot is that most White people live in predominantly White neighborhoods, and most Black people live in predominantly Black neighborhoods. Since this fact is well documented, I won’t belabor the point here.[3]

As I noted above, people who inhabit a shared environments tend to converge on similar psychological and behavioral dispositions, and they learn to devote attention to phenomena that are salient to the stability of their shared material and social environments. This allows forms of socialized attention to arise, as behaviors that conform to cultural rules are socially reinforced, while behaviors that contradict local norms are punished and abandoned (Kitayama, Park, & Cho 2015, 86). But even where sanctions are not imposed explicitly, the structure of our personal interactions with the members of other groups can shape what we pay attention to, and how we evaluate the things that we experience. Many people who inhabit White spaces never interact with people who are racialized in other ways; and often those who do, assume that the Whiteness of their space makes it safer and cleaner than other kinds of spaces. In part, this is because their experience of race is filtered through the distorting lens of mainstream media, and through encounters that are easily categorized as interactions with ‘coworkers’ or ‘acquaintances’ (not as encounters with African-Americans, Salvadorans, Kenyans, and Ethiopians). So if racist ideologies foster initial forms of biased thinking, these can solidify because the only contradictory data comes from interpersonal interactions that occur in a narrow range of social situations (e.g., only interacting with people of color in service-work contexts). Over time, stereotypes and biases can come to dominate everyday thinking about race for the inhabitants of White spaces; and the hegemony of White ideology can begin to infuse the attentional strategies of people who inhabit White spaces; put bluntly, it can become difficult to the inhabitants of White spaces to see Whiteness as anything but a necessary background against which to frame their thoughts.[4] The upshot is that White spaces aren’t just material structures. They are ideological and institutional structures, which are reflected in the things that White North Americans tend to read, watch, and think about. Whiteness structures the images that are consumed, the knowledge that is produced, and the institutions that structure social interactions.

As Wendy Leo Moore (2008) argues, there are also several overlapping social forces that interact to produce and stabilize White Institutional Space. People of color have routinely been excluded from positions of social and institutional power, leading to the development of educational practices that align with White values and interests; these educational structures provide a shared background against which inferences and arguments can be evaluated; and as such inferences become entrenched in explicit laws and normative frameworks, they can begin to fade into the background, and take on an apparently neutral and impartial character. Once this happens, distinctively White forms of culture can replicate through the repeated and reciprocal interactions between people and social structures, including: the distribution of social and material power along racial lines; hidden signifiers of White power and privilege that impact the character of the interactions in particular spaces; practices of justification and explanation that are grounded in racialized ideology, and racist discourse; and legal and normative frameworks that serve to protect White interests (Moore 2008).

The conceptual and normative frameworks that are produced by living in White spaces and adapting to forms of White ideology includes a largely tacit assumption that race will only be mentioned in the context of racial or ethnic minorities; this “elevates whiteness to the status of human standard, against which responses of racialized others constitute deviations that require explanation” (Salter & Adams 2013, 787). Likewise, everyday practices solidify around ‘colorblind’ ideologies, which strive to promote practices, decisions, and policies that make no reference to race. Many White Americans prefer discussions of education, employment, and economic opportunities that focus on individual characteristics, and make no reference to race; but people who accept this colorblind ideology tend to perceive less racism in the world, and tend to “indicate less support for anti-racist policy when colorblind ideology is salient” (Salter & Adams 2013, 786). Moreover, this ideology produces self-insulating forms of ignorance (Bonilla-Silva 2003; Crenshaw 2010; Mills 2007).

This ideology also sustains the assumption that White spaces are clean, safe, normal, unmarked space; and it can foster a tendency to assume that deviations from Whiteness are dirty, dangerous, and chaotic (Watt & Stenson 1988). In White spaces, the Whiteness of one’s neighbors is unremarkable, so it is easily ignored. When Whiteness becomes invisible, racial differences are more easily noticed, and White people will often assume that people who are racialized in other ways pose a threat to the safety, cleanliness, and familiarity of their White spaces. The invisibility of Whiteness can also trigger feelings of anxiousness when passing through multiracial or non-White spaces (Garner 2007, 44-45), as well as increased hostility toward people who are not White (Garner 2007, 148-149; 160-161). Importantly, this can also happen on a smaller scale, as it does when a racially diverse shopping space is perceived as dirty, disorderly, and ugly, rather than as a space that has been adapted to the local needs of a diverse population (Campkin forthcoming).

While it is not universal, and while there are spaces where such patterns do not emerge, it is striking, to me, how pervasive these material and ideologically structured forces are in the production and maintenance of White spaces. To operate a successful restaurant in Washington, D.C., for example, a restaurateur must attend to the racial composition of their establishment.

The vast majority of Census tracts on the east side of D.C. are overwhelmingly black, while the vast majority of western tracts are overwhelmingly white. Many of the tracts in southeastern D.C. have less than 1 or 2 percent white residents, while many western ones are less than 5 or 10 percent black, with very few Hispanic or Asian-American residents either (Blake 2015).

So in one sense, it shouldn’t be surprising that most of the restaurants in the city are highly segregated. But in a recent article in the Oxford American, the journalist and restaurant critic Todd Kliman (2015) reported that he had visited 160 restaurants over the course of 4 months, ranging “from ambitious fine-dining to so-called ethnic mom & pops”. In his experience, 90/160 of the restaurants (56%) had an exclusively White clientele, while only 8/160 (5%) had more than ten identifiably Black customers. According to Kliman (2015), several D.C. restaurateurs also acknowledged a widely shared, though largely tacit understanding that if less than 60% of a restaurant’s patrons are White, the space will become a predominantly Black space, as White people tend to avoid restaurants where more than 40% of the clientele are people of color.

One intriguing hypothesis is that these patterns emerge because White people tend to overestimate the number of Black customers in a restaurant. And where more than 40% of the patrons are Black, people who have attuned to the statistical structure of White spaces may feel like they are in a predominantly Black space. This would sit well with the recognition in design and marketing that white space can prioritize content, making important items pop out, because stark contrasts are significant to most human brains, and we use them to distinguish signals from noise. In design, white spaces fade into the background, and are easily ignored. In line with the socialized attention hypothesis, racial segregation may increase the salience of racial differences to ongoing processing by making people of color may pop-out for those who primarily inhabit White neighborhoods, and primarily pass through White spaces (and we already know that Whiteness tends to fade into the background for such people).

This hypothesis gains some support from data suggesting that White people routinely overestimate the number of non-White people living in their communities, or passing through the same spaces as them. For example, Bridget Byrne (2006) found that White inhabitants of London tend to overestimate the number of people of color in various social spaces; several of her informants report experiencing Brixton as a predominantly ‘Black Area’ (though census data suggest that approximately half of the inhabitants of Brixton were White, at least though the late 00s; see Marsh 2016). One of them even reports frequently realizing that she was the only White person on the street in Brixton (Byrne 2006, 102). Another informant, a school governor, reports seeing only three Caucasian children in a particular class, but on a second look, she realized that the class was about half White (Byrne 2006, 102). Finally, a woman who was searching for a ‘good school’ for her children reports going to an ‘inner city’ school and seeing that it was “probably maybe 80/85 per cent Afro-Carribean, and…maybe 5-10 per cent, 10 per cent maybe, Asian and a few other minority groups” (Byrne 2006, 125); but in actuality, 40% of the students were White (Byrne 2006, 127). These data are inconclusive, as they cannot distinguish between a shift in perceptual experience, a shift of attention that leads to an excessive focus on people who don’t look White, and a memory effect that’s driven by the fact that White people are socialized to prefer predominantly White spaces. But what is clear is that White people worry that spaces that are not majority-White are dirty, dangerous, or at the very least ‘not for them’.[5]

The social attention hypothesis would explain these psychological phenomena by appealing to long term patterns of cultural attunement. And I think that this is right; but these kinds of diachronic shifts in cognitive processing call for an explanation. And as I’ve argued elsewhere, the relevant kind of behavioral and psychological plasticity is likely to be common among embodied, mobile, and social organisms (Huebner 2013, 2016, in press). In the next section, I turn to these learning algorithms, before returning to questions about culturally situated experiences in the sequel.

4. Error-driven learning

As anyone who has spent time in a city will know, rats approach anything that has a chance of being food. But they quickly learn to avoid locally salient toxins, and foods that have a high chance of making them sick. And in laboratory contexts, rats will rapidly learn to avoid foods that are associated with sickness; but it will take them much longer to learn to avoid foods that are accompanied by electrical shocks (Garcia & Koelling 1966). Evolutionary trajectories leave marks on biological systems, and shape the things that organisms will find it easiest to learn about. Over countless generations, rats that could rapidly learn to avoid toxins were more likely to reproduce than those who could not; so most rats have an evolved tendency to learn which toxins are common in their local environment. By contrast, electrified food bowls are a novel threat for rats, and there are few enough Skinner boxes in the world for it to be beneficial for rats to develop a rapid learning algorithm for tracking shocks. In this respect, humans are a lot like rats. Like most other mammals, humans are ‘biologically prepared’ to learn about the risks, threats, and rewards that were relevant to the survival of our ancestors (Cummins & Cummins 1999; Rescorla 1988; Seligman 1971). But these basic learning tendencies stretch well beyond aversions to toxic food, dangerous predators, and threatening pathogens. Most humans can track group membership on the basis of minimal cues (Dunham et al 2011; Cikara & van Bavel 2014). Where large numbers of people agree, people tend to display a bias toward conformity when decisions are made under conditions of uncertainty(see Morgan & Laland 2012 for review; See Klucharev et al 2009 and 2011 on the potential mechanisms guiding conformist-choice). And most humans display evolved capacities for sociality and imitation that open up an enormous terrain for social learning (Henrich & McElreath 2003; Tennie et al 2009; Csibra & Gergely 2009; Zaki & Mitchell 2013).

Like most mobile organisms, a typical human possess capacities for self-corrected behavioral learning. As we move through the world, core regions of the brain extract and evaluate multiple sources of information that are salient to our current needs and interests. For example, we employ mechanisms that allow us to rapidly encode representations of the risks, threats, and rewards that we have encountered as we have moved through the world; and we can use these representations to evaluate the options we face when we must make decisions, and when we must act (Crockett 2013; Cushman 2013; Glimcher 2011; Huebner 2016; Montague 2006; Schultz, Dayan, & Montague 1997; Seligman et al 2013).[6] Each of these systems relies on a form of error-driven learning, where initial hypotheses are tested against incoming streams of data, and adjusted whenever things go better or worse than expected; and each of these systems attempts to minimize errors in its computations by making corrections in the required direction until the receipt of rewards and punishments occurs precisely as predicted. As long as the rate and value of risks, rewards, and threats are stable, these systems will produce viable capacities to evaluate current and counterfactual situations (Kishida et al 2016; Railton 2014), and to compare things that we value in different ways.

Neuroscientists have long known that even highly routinized tasks require integrating numerous inputs from different computational systems (including goal-based representations, perceptual inputs, and ongoing evaluative feedback) to yield representations that can be encoded for use by motor systems, control loops, and emulator circuits (Akins 1996, 354; Mahon & Caramazza 2008). The evaluative systems I have been discussing are part of this complex network of processes, and I contend that they generate an internal constraint on the information that we seek out, the information that is readily available, and the information that we will rely upon in deciding what we should do, and when we should do it. These evaluative systems facilitate decision-making under conditions of risk and uncertainty, and they generate the motivational “umph” that’s required to get us moving and to keep us focused on phenomena that are salient to our needs and interests (cf., Huebner 2012; Klein in press); and they organize our psychological and behavioral dispositions (Cushman 2013; Huebner 2016).

Focusing on these systems puts affect at the core of human cognition; and it suggests that we will always be in affective and evaluative states, no matter what is happening, and no matter how neutral our affective states feel (Lindquist 2013, 361). This matters, because the affective and evaluative state we are in when we encounter a risk, a threat, or a reward will have a significant impact on how we interpret its salience, and what new information we encode for future action. Hungry people are more attentive to the food that they see, and this enhances their food-related memories (Talmi et al 2013); people who are nervous about encountering snakes will tend to see them everywhere on a trail (Machery 2016); and social stress can evoke either increased levels of risk aversion, or increased impulsiveness, depending on what has happened recently. More generally, evaluative expectations are context sensitive, and they are continually adjusted in light of the current salience of risks, rewards, and threats. This has important, but unsurprising philosophical implications: The amount of data we need to confirm an evaluative hypothesis depends on how serious the implications of false positives and false negatives are; the stakes matter when we are trying to figure out whether something is a risk, a threat, or a source of benefit; and these neural systems adjust the precision of their estimates of risk and uncertainty in light of their current state. However you put the point, the upshot is that variations in expectations about the likelihood of particular outcomes can affect the significance of uncertain perceptual information.

But there is a catch, here. Social processing is not a one-and-done affair. We constantly compute evaluative representations, and we do so in ways that draw on our past experiences, but there do seem to be some types of top-down effects in processing (cf., Shea 2014)  As I noted in the first section, the brain is a hierarchically organized system, with processes that are deeper in the processing stream manipulating more abstract and conceptual representations. As Jay Van Bavel and his colleagues have shown, even locally induced forms of social motivation can shift patterns of categorization, causing a top-down effect on the depth of processing. Initial reactions allow for a quick-and-dirty evaluation of a social stimulus, subsequent processing allows for more subtle forms of categorization, as well as the emergence of ambivalence—and the resulting signals can up-regulate or down-regulate the salience of available information. Van Bavel & Cunningham (2010, 262), for example, argue that “the automatic activation of prejudiced representations and biased processes leads to discriminatory behaviour unless controlled processes driven by goals and motivations attenuate these biases”; however, evaluative responses and social motivations can be employed to moderate our immediate responses—and we can sometimes do so successfully so long as we possess the time and the motivation to do so. But even where this doesn’t occur, there is typically a back and forth interplay between more automatic and more controlled processes, which is reflected in iterated attempts to settle on responses to socially significant stimuli (Van Bavel & Cunningham 2010, 242).

What this means is that we typically begin from evolved or socially entrenched expectations, and then build upon them with iterative reprocessing. Our initial evaluative responses provide a skeletal structure for processing information about potentially salient stimuli. But while these expectations anchor processing, they aren’t the whole story about what we will do, or what we will think. Multiple expectations are often computed in parallel with approach-avoid motivations; these motivations are rapidly conceptualized as socially salient action-values, triggering forward-looking trains of thought; these thoughts then flow downward through our conceptualizations, shaping our motivations, and affecting the state of our bodies. If we know that we have reason to do so, we can revise our initial assumptions; however, we often adjust insufficiently, as we stop searching for information when we feel like we have reached a plausible value. But social information can shift the depth of processing (likewise with conceptual representations, and our current needs and interests). Consequently, the representations produced by the brain at any one time will draw on “multiple brain systems, and feature the recursive interaction between bottom-up cues (e.g., skin color or hair length) and top-down cues (e.g., attention or motivations) that interact in cycles until the evaluative system settles on a representation of a target” (Van Bavel, Xiao, & Hackel 2013, 112).

All told, the values that we assign to various possibilities are shaped by our learning histories, aspects of the world we have recently encountered, the nature of our current situation, and the current state of our body. This shouldn’t be a surprise. Being a successful agent requires adapting to the situations we are chronically immerse in (Dasgupta 2013, 271). This is the reason why we should expect patterns of convergence in evaluations where risks, threats, and rewards are relatively similar; and where aspects of the world are frequently encountered (Railton 2014). And it’s the reason why we should expect stable patterns of evaluative processing to emerge where internal states remain relatively stable; in such situations we should expect internal states should to play a more significant role in shaping behavior than contextual factors (note, this is a valuational analog of the suggestion by Friston et al [2015] and Clark [2015] that systems can minimize errors in their predictions either by changing their own state or by changing the state of the world; cf., Huebner & Glazer 2016). Our responses to the world should tend to develop in ways that make us sensitive to the demands of the situations that we typically encounter, and we will become sensitive to the local stabilities that emerge in our own internal states. Nonetheless, structural, material, and social constraints can lead different people to encounter different aspects of the world, in different affective states, with different prior expectations. And where this happens, people will end up with different representations of their evaluative landscape. This fact has interesting implications for what we notice, and what we think matters. And to see why, it will help to consider some microaggressive acts that seem trivial to many denizens of White spaces.

  • 5. Are you ready for some football?

I’ve lived in Washington D.C. for the past seven years. There is a great deal that I love about this city; but there are things about it that I find deeply puzzling, and deeply disturbing. For example, the American Football team here has a racial slur for a name, and it uses a caricature of an Indigenous person for a mascot. Throughout Washington D.C., Southern Maryland, and Northern Virginia (hereafter, the DMV), people routinely use the word ‘Redskin’ as though it were nothing more than the name of a Football team. Numerous Indigenous groups have called for an end to the use of this racialized term and the accompanying mascot (National Congress of American Indians, 2013). And in recent years, they have received increasing support from civil rights groups, politicians, and public figures. But people across the DMV, including the current owner of the team, wonder what the problem is supposed to be. To them, it’s just a name, and it doesn’t have any pejorative connotations. But at the same time, those who argue that the team should keep its name and mascot often suggest that they are honoring the virtues and achievements of Native Americans; and that those who are offended by its use should just lighten up. Many people in the DMV acknowledge that the word serves as a reference to the skin of Indigenous people (though they probably don’t entertain the possibility that it was used to mark the practice of collecting the bloody scalps of Indigenous people for bounty). And it would be hard to deny that the word functioned as a pejorative term throughout the late 19th and early 20th centuries. But the pejorative use of the term seems to have faded; and the people who use it tend to see it as a reference to a team that they love and support.

From a perspective that looked at the name and mascot of the Washington Football team on their own, without thinking about their role in larger structures of exclusion and marginalization, it might seem obvious that any experienced offense is an over-reaction. So it should come as no surprise that the conversations that have developed around changing the name of this team have paralleled the recent theoretical discussions of microaggressions. The people who are demanding that the name and mascot be changed seem to be calling for public acknowledgement of a minor offense, and for a large scale change that vastly outstrips the nature of the offense. Perhaps this strikes some people as evidence of an emerging culture of victimhood (Campbell & Manning 2014), or as evidence of the deep social entrenchment of psychological fragility (Lukianoff & Haidt 2015). But it’s important to remember that when people call attention to microaggressions, they typically do so as a way of highlighting larger patterns of exclusion or marginalization; and in this case, the relevant patterns include a history of genocide and ethnic cleansing, as well as robust contemporary stereotypes about Indigenous Americans.

When a man comments approvingly on another man’s clothing as he’s walking down the street, or suggests to a White colleague that he should be less aggressive, we rarely find fault with his comments. But if the same man comments approvingly on a woman’s clothing as she walks down the street, or tells his Black colleague that he should be less aggressive, these acts can take on a decidedly different character. Likewise, against the backdrop of structural racism directed toward Indigenous people, the mundane act of referencing a Football team becomes microaggressive; and it is thereby likely to play a significant role in producing and perpetuating patterns of exclusion, marginalization, and disadvantage (cf., Leibow in prep).

Microaggressive acts often occur with little or no recognition that they are problematic, even though they often reflect patterns of internalized bias. And if we focus on the mindset of the individual who acts microaggressively, it will be easy to assume an individualist and psychological perspective, which focuses on the outputs of associative processes, behavioral manifestations of implicit beliefs, or expressions of biased dispositions (cf., Slater & Adams 2013). Adopting this approach, however, threatens to downplay one of the most intriguing facts about the harms generated by microaggressions. Microaggressions cause harm because they are constituents of spatially and temporally distributed macroaggressions, which are carried out by multiple people in multiple places; put differently, aggregated behaviors can yield aggressive acts, even if individuals do not, and perhaps cannot perceive the aggressiveness of their individual actions. This is because each microaggression is part of a larger social structure, which is constituted by numerous, interrelated, but physically and temporally distinct expressions of bias and animosity. The nature of these aggressions would be clearer to observers if they were carried out by a single individual; and when they are distributed across multiple spaces and times, it becomes hard to see the unity among their diversity. That said, there are ways of compressing these networks of spatially and temporally distributed aggression to make it clear how individual acts are unified by a distributed power structure.

Consider a case study, which brings the relationship between microaggression and macroaggression into stark relief. On 27 September 2014, Gregg Deal, an Indigenous artist (Pyramid Lake Paiute), organized a performance art piece entitled “REDSKIN”, as part of a growing movement to change the name of the American football team. This piece was designed to illustrate the microaggressions that are routinely experienced by Indigenous Americans, and to clarify their role in sustaining cultural norms of exclusion, marginalization, and oppression (for an overview, see http://greggdeal.com/REDSKIN). No one knew quite what to expect when they entered the shell of a gutted convenience store where the performance took place; but what the attendees experienced was a powerful representation of the micro- and macroaggressions that are routinely experienced by Indigenous Americans. For approximately 4 hours (two performances of roughly two hours each), Deal sat silently as four associates (including the current author) hurled microaggressive speech acts at him, mocking his heritage, and reducing him to a stereotype. Instead of using a script, the piece built on a wide variety of comment threads from news stories about the attempts to change the name of the Washington football team; so the performance was improvisational, and the content emerged as the actors played off one another’s racialized speech. On their own, most of the speech acts have would have proven easy for skeptics to dismiss. But in the context of this performance, it was hard to ignore the racialized power present in each of them.

From the perspective of someone who has become aware of the larger structures of oppression and exclusion that are experienced by Indigenous people, it is hard to understand why debates continue over the name of the Washington football team. It is a racial slur; and it is part of a network of systematic exclusion and disadvantage. Particular uses of this slur, and particular representations of Indigenous people, are carried out by particular individuals; likewise for other acts of demeaning and marginalizing rhetoric, which are distributed across many different actors, many of whom have never met one another. But together, they cause a great deal of harm and offense, at least in part because of the roles that they play in larger networks of institutionalized social practices. People who have not encountered this network of ideological practice may have a difficult time understanding why the microaggressions experienced by Indigenous people are problematic. But by compressing a wide variety of slurs and insults into a tightly organized artwork, Deal demonstrated that macroaggressive practices can emerge through microaggressive behaviors. Few of the observers that night would have experienced the distributed structure of this aggression in the context of their everyday lived experience. But in the context of the performance piece, the sense of oppression and marginalization was palpable, and at least one person left with a much better sense of how racialized power flows through social groups.

Having taken part in this performance, I find it hard to understand what is going on when people suggest that discussions of racial microaggression are misguided. It is often suggested that the focus on trivial and unintended offenses will distract us from the project of addressing “real aggressions” (Etzioni 2014). And it is sometimes claimed that such discussions will discourage people from engaging in genuine and spontaneous interactions, license unjustified forms of anger in response to minor infractions, and foster a mentality of victimhood instead of a mentality of openness (DeAngelis 2009; Chait 2015). They may even seem to compromise free academic inquiry by leading people to demand public recognition of even minor and unintentional offenses (Haidt 2015; cf., Campbell & Manning 2014). But from the perspective of those who are frequent targets of microaggressions, things look very different. The arguments advanced by critics seem to ignore the role of social power in maintaining spaces that are structured by historical and contemporary injustice; by calling attention to microaggressions, many people hope to: “build power in communities that overwhelmingly experience such aggressions” (Mukhopadhyay 2015) while also providing concrete reminders that some people experience whole classes of people as ethnicities or genders. And just as importantly, some people hope that offering reminders to use respectful language can have “a positive effect on how people perceive each other and themselves” (Mukhopadhyay 2015).

Each of these responses probably feels plausible from the perspective of someone who advances it. And each of them probably feels ungrounded from the perspective of someone with a different learning history, a different form of embodiment, or a different social location. This doesn’t mean that these perspectives are equally plausible; but each of them is intelligible as an initial response that is likely to emerge from a particular location in social space and a particular learning history. Thinking about these responses through the lense of cultural psychology and evaluative learning may open up a novel perspective for understanding how these differences arise between people who routinely experience racialized hostility and people who encounter microaggressions only as spectators, or occasional perpetrators. As I noted above, the psychological and behavioral tendencies of people who inhabit similar environments will tend to converge; but when people inhabit different environments, “that are configured with ideas, practices, and institutions that do not construct the self as the primary source of action, strikingly different psychological tendencies are revealed” (Markus & Kitayama 2010, 428).

People who inhabit positions of lower social status are likely to focus more attention on contextual factors, including the structural and institutional causes of systematic disadvantage (Kraus et al 2012). The effects of institutional power impinges on their bodies frequently, in a variety of different situations; and they all hang together as modes of concrete and material oppression. If the story I’ve been telling is approximately right, people who routinely experience these forms of social and individual contingency should often find it easier to perceive the networks of macroaggressive power that flow through microaggressive speech acts. There are a couple of reasons for this. First a person who anticipates racial hostility will be more likely to notice the racial connotations of a single action; this expectation should also lower their threshold for identifying an act as a microaggressive. This is something that critics may acknowledge. But second, and just as importantly, these evaluative systems can track socially stable but spatially and temporally distributed regularities; and each experienced microaggression can be embedded in a larger internal model of the evaluative landscape (e.g., one that represents the world as structured around racialized forms of power). Where this happens, this may yield a shift in attention, away from individual actions, and toward the patterns of exclusion, marginalization, and oppression that are stable features of the social environment. This is not to say that the individual actions won’t be tracked, or that they won’t be salient. My claim is only that the social and institutional structure of microaggressions, as well as their roles in larger macroaggressions, should be more salient to the interpretation of an action for a person in this context (though perhaps only sub-personally; I don’t know how salient these representations are at the person-level). So long as the relevant social structures exist, it should be possible to track them; and the variations between the individual actions that express these macroaggressions should be less salient to ongoing behavior than the stability of the pattern itself.

By contrast, people who inhabit social secure class positions are likely to attune to the role of internal influences on behavior, heightening their concern with personal freedom (Kraus et al 2012, 550), and leading them to neglect the structural factors that organize and perpetuate patterns of oppression, exclusion and marginalization. Such positions tend to yield atomistic and individualistic conceptions of racism, which “have a foundation in broader conceptions of self and society that locate action and experience in isolated individuals abstracted from social context” (Salter & Adams 2013, 785). But these attitudes are not neutral, rationally-basic, or objective; they are grounded in evaluative responses that “reflect the particular experience of people—most prototypically, propertied White men—whose identity positions (and socioeconomic correlates) afford them the experience of abstraction from context” (Salter & Adams 2013, 785). If this is right, positions of high social status will tend to foster a focus on individual traits. Someone who experiences a pattern of racial hostility primarily as a spectator will also be less likely to notice the racial connotations of a single action; and their threshold for identifying an act as a microaggression will tend to be much higher. Without a clear sense of the patterns of structural exclusion that microaggressions represent, calls for redress will seem unsuited to the minor offenses that they are directed upon. And since microaggressions are not experienced in states that highlight them as significant features of local environments, they will tend to be experienced as expressions of the internal states of individual actors.

As I noted above, these will only be tendencies, though they are likely to be relatively stable in the world that we now inhabit as quick-and-dirty responses. That said, even these patterns of regularities are likely to be complicated by biologically-grounded variations in risk aversion, impulsivity, reward sensitivity, and perceptual acuity, all of which can impact the particular features of the world that a person will attend to. These patterns of biological variation can yield complex networks of differences in our understandings of what is possible, and this is why even the most hegemonic power structures are likely to be contested.[7] Likewise, differences in the precise trajectories that we take the world can affect the patterns that we attend to, and yield relatively stable individual differences in evaluative tendencies. But it is important to remember that we take part in social and cultural practices, we don’t just observe them. These are not just ways of thinking, they are ways of acting in the world, and staking out our positions in social space. As I noted at the outset, cultural attunement requires active, repeated, and ongoing participation in culturally scripted behavior (Park & Kitayama 2011, 77). And as we participate in culturally structured practices, we replicate the forces that impose structure on our social environments, often by imposing additional pressures on one another to conform. We are social niche constructors, and “socioculturally shaped shapers” of our environment (Markus & Kitayama 2010, 421). Through social nudges, bumps, and hailings, we play an active role in shaping the possibilities that others experience as available. To see what this  means, let’s return once again to Gregg Deal’s performance piece.

6. Situational adaptation and contestation

When it was time for the performance of REDSKIN to begin, I found it difficult to know how to get started. I knew what people tend to say to offend Indigenous people; after all, I had read numerous comment threads on blogs and editorials, and I knew which comments were ‘right’ for this piece. But like most people in the United States, I have been socialized to avoid saying things that I explicitly recognize to be racist in public. This is a weak constraint, which makes quite a bit of room for coded speech, dog whistles, and ambiguously racist claims; but this performance required me to say things precisely because I recognized that they were racist, and this evoked a form of cognitive conflict that was initially hard to overcome. But it was troubling how quickly these difficulties dissipated, and how quickly my evaluative state adapted to the context I was in. Among friends who were saying racist things, the words started to come more easily. As performers, we built off of one another’s actions, and we rapidly constructed a stable micro-world where we were not sanctioned for racist speech. As I listened to the other performers that night, I began to feel more comfortable in my role in the performance. This is the predictable, but deeply problematic effect of the capacity for socially-situated and adaptive evaluative learning. In this context, my evaluations were rapidly calibrated against the structure of my local environment, which was an environment where racist speech was accepted and necessary; I was in the flow, and this was a bad thing.

A similar form of local adaptation happened to the audience, but in a way that helps to highlight the possibility (as well as the fragility) of local forms of contestation. The performance was part of “Art All Night” (a free art festival that includes visual art, theater, dance, performance art, and music). People knew that the space where we were performing included art that was focused on the name of the Washington Football Team, and the murals Deal had painted provided clear signals about the kind of exhibit they were seeing (See http://greggdeal.com/REDSKIN, especially images 7 & 10). But as I noted above, the members of the audience didn’t have firm expectations about what kind of performance they would see, and they didn’t know if they would be looking at art on the walls, watching a performance, or experiencing something else all together. The first audience watched silently; and their silence became powerful, as it brought the performance to center stage. But as we started the second performance, someone in the audience started to chant: “Change The Name! Change The Name! Change The Name!”. The sentiment resonated through the audience, and soon others joined in; but this low-cost form of resistance didn’t last long. The chant died down. The performance continued. And the sense of racial animosity once again began to permeate the room…

This brings us to a second and final layer of cognitive processing. We acquire the capacity to represent many aspects of our evaluative landscape reflexively, but in our actions we also call upon others to aid us in the construction of shared micro-worlds (Kukla & Lance 2009). In social situations, our focus will often shift to finding ways to bring or behavior into alignment with the people around us; and as we interact with one another, the resulting social context can produce new understandings of the options that are available to us (Gallotti & Frith 2013). This is often a good thing. We couldn’t carry out interpersonal transactions if our patterns of thought and action weren’t aligned with the thoughts and actions of people around us, and we use social feedback to shape our reflexive evaluations in ways that make us seem like good collaborator (cf., Slingerland 2014); but in contexts that are organized by racial power, people will tend to converge on racist ideologies and practices, as they will tend to receive interpersonal support for these attitudes. I found it easy to engage in microaggressive practices because my friends were doing so, and I rapidly adapted to the local structure of a micro-world where it was acceptable to knowingly express racist ideas; and a single audience member who spoke up, opened the door to a more general rejection of that micro-world.

Unfortunately, attempts to construct shared expectations that run contrary to dominant social norms often give way to actions that are more aligned with the norms and practices we are trying to overcome. This happens because actions driven by counter-social motivations are met with interpersonal feedback suggesting that we are making normative mistakes; and this feedback triggers the production of error-signals in our evaluative learning-systems, because we tend to treat conformity with local norms as intrinsically rewarding, and deviance from local norms as errors to be corrected (Klucharev et al 2009, 2011; Huebner 2016). Since practical competence is guided by evaluative expectations that are adjusted when, and only when, things do not go as expected, our biases often become calcified in social practices like those I have been discussing. Where we are attuned to biased practices, our ongoing behavior will help to entrench problematic practices, leading to more robustly biased structures against which our future attitudes will become attuned. And when the vast majority of our intuitions are attuned to a messed up world, we are likely to rely on problematic assumptions about what’s right and what’s wrong, as well as what counts as evidence for and against our reflective hypotheses; and where the sense that we are making normative mistakes continually arises, we will adjust our behavior to make it more consistent with the local micro-worlds that we inhabit. Put much too simply, as the brain searches for the linked set of hypotheses that make incoming data most plausible, our expectations will shift toward statistically common patterns that we hope to overcome (Huebner 2016).

Moreover, since we tend to reason from a position within existing statistical and ideological structures, the options that we treat as open and available often depend on the questions that others lay out for us, and the norms for responding to them that we are attuned to. As a result, our language games come to constitute another statistical regularities for us to attune to, and as such they come to guide the unfolding of our socially relevant actions. For example, in the United States, White children are typically taught that they should not talk about race. And both explicit social ideology, as well as implicitly learned evaluative tendencies, lead them to conceptualize “colorblindness” as an ideal to be pursued. However, their evaluative expectations will often reflect worries about the safety, cleanliness, and quality of social spaces (e.g., restaurants, neighborhoods, and schools) where White people are the minority. This kind of cognitive conflict is likely to evoke deeper forms of reprocessing, with an attempt to minimize the cognitive dissonance that results from these conflicting pulls. Something similar happens in cases where people feel the need to preserve a non-racist identity, while also feeling the need to categorize someone in terms of their racilized identity. To navigate these forms of cognitive conflict, groups of people will often develop coding strategies, which allow them to feel more comfortable with their patterns of speech. In the context of casual racism, the use of words like ‘ghetto’, ‘thug’, ‘urban’, ‘inner city’, ‘dangerous’, ‘welfare’, ‘food stamps’, and ‘illegal immigrant’ can help to minimize cognitive dissonance, as it is possible to dissemble from racial connotations if doing so is necessary. Nonetheless, everyone who has been socialized in a community where these words stand-in for racial content will know exactly what these terms are meant to convey (cf., Clifton 2015; Haney-López 2015; Savali 2015). But I think that the reason for this may be philosophically and psychologically interesting.

A recent perspective in cognitive neuroscience opens up a way of thinking about this kind of process, in a context which is tied to the kind of evaluative cognition that I have been discussing thus far. Many humans possess capacities to triangulate their evaluative representations against the understandings of the world that others possess. There are two forms of metacognition accessible to most humans, “individual learning” kinds (which are commonly shared with other animals), and broadcast-and-revision kinds (which are probably distinctively human). We rely on the latter kinds of processing to select “metacognitive information for broadcast, in the service of controlling the sensorimotor systems of two or more agents involved in a shared task—that is, for supra-personal cognitive control” (Shea et al 2014, 189). Consider the difference between training an intelligent nonhuman primate to participate in a simple experiment, and asking an undergraduate student to participate in a similar task. The nonhuman primate will have to be trained and reinforced from the bottom-up; but the undergraduate will often immediately acquire an understanding of the task as soon as they hear the question (Roepstorff & Frith 2004). Unsurprisingly, this difference runs primarily through our competence with language. We can broadcast linguistically-structured representations in ways that shape the expectations that others form, and this opens up the possibility of using a supra-personal kind of cognitive processing to control how social evaluations and socially significant actions unfold (Shea et al 2014; Zawidzki 2013).

Whatever else racist code words do, they also play a significant role in bringing our evaluative expectations into alignment. For example, if I say that a neighborhood is a little bit dangerous in Washington DC, I will implicitly flag that it is a predominately Black, or a multi-racial neighborhood. In some situations, this will allow me to send a compressed signal which can be unpacked (by some observers) in ways that will shift attention from the new bar that is about to open there, to the worries about being threatened or harassed by a non-white assailant. Of course, for this to work, a sender and a receiver have to share evaluative assumptions, and they need to compile the signal in similar ways; this doesn’t always happen, and such signals can fail to hit their mark (e.g., the recently expressed trumpish worries about taco trucks on every corner). But where patterns of speech solidify and stabilize, the process of compiling and decompiling can become habitual and automatic; and as we attune to these patterns of interpretation, this will tend to generate and sustain racialized perspectives. Put bluntly, code words can provide the conceptual structure against which we calibrate our picture of the world, and they come to shape the options that we perceive as available, as well as the options that we perceive as off the table. And they often help us to set up shared social spaces, where we can think together, in racial terms, without compromising the thin veneer of colorblindness.

7. In lieu of a conclusion

Biological and social cognition requires sifting through and prioritizing a massive amount of potentially relevant information, doing so in a way that is sensitive to the difference between better and worse options, and updating our assumptions about what is better as our options change. Only a small fraction of the information that we encounter and process will ever be relevant to our current and ongoing concerns, and how important it is will always depend on our current situation as well as what else has happened recently. For most purposes, conscious and deliberate thinking is too slow and computationally expensive to do the job that is required. So like other animals, we often rely on evaluative systems that are sensitive to the distribution and value of rewards, the probability of gains and losses, and subjective estimates of risk and uncertainty. These mechanisms compute ‘predictions’ about what the world is like, and they motivate thought and behavior in line with these predictions; but they also update future predictions in ways that minimize discrepancies between ‘predicted’ and actual outcomes. Over time, where the structure of the world is fairly stable, these ‘predictions’ will yield accurate representations of the world. By way of error-driven learning, we attune to the distribution and value of the risks, rewards, and opportunities we are likely to encounter.

As socially situated animals, who inhabit socially structured echo chambers, it’s often difficult for us to determine which of the patterns we perceive are locally accurate but globally mistaken; and we often infer that the local patterns we track accurately are global patterns that everyone experiences. Put somewhat more figuratively, the cognitive maps we construct as we pass through an evaluative landscape are tied to the local contingencies that we have experienced. Like all maps, these ones will often represent our own trajectories fairly accurately; but like all maps, errors will tend to emerge where our evidence is speculative, and where our understanding of the world is filtered through distorted forms of cultural transmission. We can get a good idea of where our more general inferences are right, by finding places where interpersonal similarities emerge. And as I’ve just suggested, we can coordinate across individual differences, as well as differences in moods, expectations, and biases; but our success in doing so varies as a result of our openness to others, and our willingness to reconsider our initial construal of a situation. But so long as we interact with people who are passing through similar cultural and material trajectories as us, we run the risk of reinforcing the local regularities that emerge in locally stable micro-worlds, and treating them as evidence of global regularities. The question is, how can we move forward? I have some thoughts on these issues (Huebner in press). But for now, let me just say that addressing this issue calls for a deeper inquiry into our capacities for thinking and planning together.

8. Acknowledgments

I would like to thank Gregg Deal for asking me to take part in his performance of REDSKIN; it was an amazing opportunity, and I’m glad that he trusted me enough to let me take part in it. I would also like to thank Nathaniel Adam Tobias Coleman, for recommending Gardner’s book on Whiteness, and asking me to think critically about its contents. And I would like to thank Susanna Siegel and Nico Silins for inviting me to the NEH Summer Institute on Presuppositions and Perceptions; my four weeks at that institute provided me with an opportunity to talk with amazing people, and to try out some of the material in this paper. I presented some of this material at Virginia Tech, The Southern Society for Philosophy and Psychology (Louisville), at L’Associazione Italiana di Scienze Cognitive (Genova), The UMD Social Minds Graduate Conference, and at a conference on Sellars’ Legacy (Beirut). But the version of the ideas that I present in this paper took shape only after the NEH summer institute. Finally, I would like to explicitly thank Liam Kofi Bright, Becko Copenhaver, Cliff Grandy, Joey Jebari, Ruth Kramer, Rebecca Kukla, Eli Kukla-Manning, Jenni Mueller, Cat Prueitt, and Susanna Siegel for conversations that helped me to clarify my thinking on some of the issues in this paper. This is only the beginning of this project, and I have a long ways to go before I get all of the things that I learned from these conversations down on paper.


Works cited

Adolphs, R. (2010). What does the amygdala contribute to social cognition? Annals of the New York Academy of Sciences 1191: 42-61.

Akins, K. (1996). “Of Sensory Systems and the ‘Aboutness’ of Mental States,” The Journal of Philosophy, 93, 337-72.

Badger, E. (2016). This can’t happen by accident. the Washington post. 2 May 2016; retrieved 4 September 2016 from goo.gl/SgwDFQ

Barrett, L.F. (2015). When a gun is not a gun. New York Times, 17 April 2015; retrieved 12 August 2015 from http://goo.gl/PvXLDt.

Blake, A. (2015). The remarkable racial segregation of Washington, D.C., in 1 map. The Washington Post. 19 June 2015; retrieved from goo.gl/cQpnru on 4 September 2016.

Bonilla-Silva, E. (2003). ‘New racism,’color-blind racism, and the future of whiteness in America. White out: The continuing significance of racism, 271-84.

Byrne, B. (2006). White lives: The interplay of ‘race’, class and gender in everyday life. Routledge.

Campbell, B., & Manning, J. (2014). Microaggression and moral cultures. comparative sociology, 13(6), 692-726.

Campin, B. (forthcoming). Surviving ‘regeneration’. Think Pieces: A Journal of the Joint Faculty
Institute of Graduate Studies, University College London

Chait, J. (2015). Not a Very PC Thing to Say: How the Language Police are Perverting Liberalism. New York Magazine, 27, 12.

Cikara, M., & Van Bavel, J. J. (2014). The neuroscience of intergroup relations an integrative review. Perspectives on Psychological Science, 9(3), 245-274.

Clark, A. (2015). Surfing uncertainty: Prediction, action, and the embodied mind. Oxford University Press.

Clark, A. (2015). Strange Inversions: Prediction and the Explanation of Conscious Experience. In Essays on Dennett. B. Huebner (ed). New York: Oxford University Press

Clifton, D. (2015). There’s One Word We Need to Stop Using When Talking About the Baltimore Riots; retrieved from goo.gl/1R8DRR on 4 Sept 2016.

Coates, T. (2014). The case for reparations. The Atlantic Monthly. June 2014; retrieved on 4 September 2016 from goo.gl/JGJSRh

Correll, J., Wittenbrink, B., Crawford, M. T., & Sadler, M. S. (2015). Stereotypic vision: How stereotypes disambiguate visual stimuli. Journal of personality and social psychology, 108(2), 219.

Crenshaw, K.W. (2010). Twenty years of critical race theory: Looking back to move forward. Conn. L. Rev., 43, 1253.

Crockett, M. (2013). Models of morality. Trends in Cognitive Science, 17, 8, 363-6.

Cummins, D. D., & Cummins, R. (1999). Biological preparedness and evolutionary explanation. Cognition,
73(3), B37-B53.

Cushman, F. (2013). Action, outcome and value: A dual-system framework for morality. Personality and Social Psychology Review, 17 (3), 273-292.

Dasgupta, N. (2014). Implicit attitudes and beliefs adapt to situations: A decade ofresearch on the malleability of implicit prejudice, stereotypes, and the self-concept. In P.G. Devine & E.A. Plant (Eds.). Advances in Experimental Social Psychology, 47, 233-279. UK: Academic Press.

DeAngelis, T. (2009). Unmasking ‘racial micro aggressions’. Monitor on psychology, 40 (2), 42.

DiAngelo, R. (2011). White fragility. The International Journal of Critical Pedagogy, 3(3).

Dunham, Y., Baron, A. S., & Carey, S. (2011). Consequences of “minimal” group affiliations in children. Child development, 82(3), 793-811.

Etzioni, A. (2014). Don’t Sweat the Microaggressions. The Atlantic. 8 April 2014; retrieved from goo.gl/3MjDg2 on 5 Sept 2016

Firestone, C., & Scholl, B. J. (2015). Cognition does not affect perception: Evaluating the evidence for ‘top-down’effects. Behavioral and brain sciences, 1-72.

Friston, K., Rigoli, F., Ognibene, D., Mathys, C., Fitzgerald, T., & Pezzulo, G. (2015). Active inference and epistemic value. Cognitive neuroscience, 6(4), 187-214.

Frith, C. (2013). Making up the mind: How the brain creates our mental world. John Wiley & Sons.

Gallotti, M., & Frith, C. D. (2013). Social cognition in the we-mode. Trends in cognitive sciences, 17(4), 160-165.

Garcia, J., & Koelling, R. A. (1966). Relation of cue to consequence in avoidance learning. Psychonomic Science, 4, 123-124.

Garner, S. (2007). Whiteness: An Introduction. London: Routledge.

Glimcher, P. (2011). Understanding dopamine and reinforcement learning: the dopamine reward prediction error hypothesis. Proceedings of the National Academy of Sciences, 108 (Supplement 3), 15647-15654.

Csibra, G., & Gergely, G. (2009). Natural pedagogy. Trends in cognitive sciences, 13(4), 148-153.

Goff, P. A., Jackson, M. C., Di Leone, B. A. L., Culotta, C. M., & DiTomasso, N. A. (2014). The essence of innocence: Consequences of dehumanizing Black children. Journal of personality and social psychology, 106(4), 526.

Gross, S., Chaisilprungraung, P., Kaplan, E., Menendez, J., & Flombaum, J. I. (In Press). Problems for the purported cognitive penetration of perceptual color experience and Macpherson’s proposed mechanism. In ( E. Machery & J. Prinz Eds.), Perception & Concepts (Baltic International Yearbook of Cognition, Logic, and Communication, Vol. 9). New Prairie Press.

Haidt, J. (2015). Where microaggressions  really come from: a sociological account. Retrieved from goo.gl/p6hdPx on 5 September 2016.

Hall, M., Crowder, K., & Spring, A. (2015). Neighborhood foreclosures, racial/ethnic transitions, and residential segregation. American sociological review, 80(3), 526..

Haney-López, I. (2015). Dog whistle politics: How coded racial appeals have reinvented racism and wrecked the middle class. Oxford University Press.

Henrich, J., Heine, S. J., & Norenzayan, A. (2010). The weirdest people in the world?. Behavioral and brain sciences, 33(2-3), 61-83.

Henrich, J., & McElreath, R. (2003). The evolution of cultural evolution. Evolutionary Anthropology: Issues, News, and Reviews, 12(3), 123-135.

Hohwy, J. (2013). The predictive mind. Oxford University Press.

Huebner, B. (2013). Socially embedded cognition. Cognitive systems research, 25, 13-18.

Huebner, B. (2015). Do emotions play a constitutive role in moral cognition?.Topoi, 34(2), 427-440.

Huebner, B. (2016). Implicit Bias, Reinforcement Learning, and Scaffolded Moral Cognition. In Brownstein, M. & J. Saul (Eds.). Implicit Bias and Philosophy, Volume 1: Metaphysics and Epistemology. Oxford University Press, 47-79.

Huebner, B. (In press). Planning and prefigurative practice. In The philosophy of Daniel Dennett. New York: Oxford University Press.

Huebner, B. & T. Glazer (forthcoming). Emotional Processing in Individual and Social Recalibration. The Routledge handbook of the philosophy of the social mind. Kiverstein, J. (ed.). New York: Routledge,

Kliman, T. (2015). Coding and decoding dinner. Oxford American, 88, 13 May 2015; retrieved from goo.gl/U9TS7V on 4 September 2016.

Kishida, K.T., Saez, I., Lohrenz, T., Witcher, M.R., Laxton, A.W., Tatter, S.B., White, J.P., Ellis, T.L., Phillips, P.E. and Montague, P.R. (2016). Subsecond dopamine fluctuations in human striatum encode superposed error signals about actual and counterfactual reward. Proceedings of the National Academy of Sciences, 113(1), 200-205.

Kitayama, S., & Park, H. (2011). Perceiving through culture: The socialized attention hypothesis. In N. Ambady, K. Nakayama, S. Shimojo and R. B. Adams, Jr. (Eds.), Social Vision. New York: Oxford University Press.

Kitayama, S., Park, J., & Cho, Y.H. (2015). Culture and neuroplasticity. In M. J. Gelfand, C. Y. Chiu, & Y.-Y. Hong, Advances in culture and psychology (vol. 5). New York: Oxford University Press. http://dx.doi. org/10.1093/acprof:oso/9780190218966.003.0002

Kitayama, S., & Uskul, A. K. (2011). Culture, mind, and the brain: Current evidence and future directions. Annual review of psychology, 62, 419-449.

Klein, C. (in press). What do predictive coders want? Synthese.

Klucharev V., Hytönen K., Rijpkema M., Smidts A., Fernández G. (2009). Reinforcement learning signal predicts social conformity. Neuron 61, 140–151

Klucharev V., Munneke M., Smidts A., Fernández G. (2011). Downregulation of the posterior medial frontal cortex prevents social conformity. J. Neurosci. 31, 11934–1194

Kraus, M. W., Piff, P. K., Mendoza-Denton, R., Rheinschmidt, M. L., & Keltner, D. (2012). Social class, solipsism, and contextualism: how the rich are different from the poor. Psychological review, 119(3), 546.

Kukla, R., & Lance, M. N. (2009). ‘Yo!’and’Lo!’: The Pragmatic Topography of the Space of Reasons. Cambridge: Harvard University Press.

Leibow, N. (in prep). Microaggressions and moral responsibility.

Levin, D. T., & Banaji, M. R. (2006). Distortions in the perceived lightness of faces: the role of race categories. Journal of Experimental Psychology: General, 135(4), 501.

Lindquist, K. A. (2013). Emotions emerge from more basic psychological ingredients: A modern psychological constructionist model. Emotion Review, 5(4), 356-368.

Lukianoff, G., & Haidt, J. (2015). The coddling of the American mind. The Atlantic, (September Issue).

Machery, E. (2015). Cognitive penetrability: a no-progress report. Zeimbekis, J., & Raftopoulos, A.(2015). The cognitive penetrability of perception. New philosophical perspectives. New York: OUP.

Madrigal, A. (2014). The Racist Housing Policy That Made Your Neighborhood. The Atlantic. Retrieved from goo.gl/Pec5VC on 04/09/16.

Mahon, B. Z., & Caramazza, A. (2008). A critical look at the embodied cognition hypothesis and a new proposal for grounding conceptual content. Journal of physiology-Paris, 102(1), 59-70.

Markus, H. R., & Kitayama, S. (1991). Culture and the self: Implications for cognition, emotion, and motivation. Psychological review, 98(2), 224.

Markus, H. R., & Kitayama, S. (2010). Cultures and selves A cycle of mutual constitution. Perspectives on Psychological Science, 5(4), 420-430.

Marsh, S. (2016). How has Brixton really changed? The data behind the story. The guardian. 14 January 2016; retrieved from goo.gl/QwrmpX on 4 September 2016.

Mekawi, Y., & Bresin, K. (2015). Is the evidence from racial bias shooting task studies a smoking gun? Results from a meta-analysis. Journal of Experimental Social Psychology, 61, 120-130.

Mills, C. (2007). White ignorance. Race and epistemologies of ignorance, 11-38.

Montague P. (2006). Why Choose This Book? How We Make Decisions. New York: Dutton.

Montague, P. R., Dolan, R. J., Friston, K. J., & Dayan P. (2012). Computational psychiatry. Cognitive Science 16: 72–80.

Moore, W.L. (2008). Reproducing racism: White space, elite law schools, and racial inequality. Rowman & Littlefield.

Morgan, T. J. H., & Laland, K. N. (2012). The biological bases of conformity. Frontiers in neuroscience, 6, 87.

Mukhopadhyay, S. (2015). Stop complaining about ‘victimhood culture’. Al Jazeera America. 8 October 2015; retrieved from goo.gl/nnMZAW on 5 September 2016.

National Congress of American Indians (2013). Ending the legacy of racism in sports and the era of harmful “Indian” sports mascots. Retrieved from goo.gl/gXnMH0 on 5 September 2016

Norenzayan, A., Shariff, A. F., Gervais, W. M., Willard, A. K., McNamara, R. A., Slingerland, E., & Henrich, J. (2016). The cultural evolution of prosocial religions. Behavioral and brain sciences, 39, e1.

Railton, P. (2014). Reliance, Trust, and Belief. Inquiry, 57(1), 122-150.

Rescorla, R. (1988). Pavlovian conditioning: It’s not what you think it is. American Psychologist, 43, 151-160.

Richerson, P. J., & Boyd, R. (2008). Not By Genes Alone: How Culture Transformed Human Evolution. University of Chicago Press.

Roepstorff, A., & Frith, C. (2004). What’s at the top in the top-down control of action? Script-sharing and ‘top-top’ control of action in cognitive experiments. Psychological Research, 68(2-3), 189-198.

Salter, P., & Adams, G. (2013). Toward a critical race psychology. Social and Personality Psychology Compass, 7(11), 781-793.

Savali, K. (2015). Harry Belafonte on ‘Thugs,’ Capitalism and Women Controlling the Universe. The Root. 14 May 2015; retrieved from goo.gl/4OAjOL 4 September 2016.

Schultz, W. (1998). Predictive reward signal of dopamine neurons. Journal of neurophysiology, 80 (1), 1-27.

Schultz, W. (2010). Dopamine signals for reward value and risk: basic and recent data. Behavioral and brain functions, 6(1), 1.

Schultz, W., Dayan, P., & Montague, P. (1997). A Neural Substrate of Prediction and Reward. Science, 275, 1593-1599.

Seligman, M. E. (1971). Phobias and preparedness. Behavior therapy, 2(3), 307-320.

Seligman, M., P. Railton, R. Baumeister, & C. Sripada (2013). Navigating into the future or driven by the
past. Perspectives on Psychological Science, 8(2), 119-141.

Shea, N. (2014). Distinguishing top-down from bottom-up effects. Perception and its modalities, 73-91.

Shea, N., Boldt, A., Bang, D., Yeung, N., Heyes, C., & Frith, C. D. (2014). Supra-personal cognitive control and metacognition. Trends in cognitive sciences, 18(4), 186-193.

Slingerland, E. (2014). Trying Not to Try: Ancient China, Modern Science, and the Power of Spontaneity. Crown.

Talmi, D., Ziegler, M., Hawksworth, J., Lalani, S., Herman, C. P., & Moscovitch, M. (2013). Emotional stimuli exert parallel effects on attention and memory. Cognition & emotion, 27(3), 530-538.

Tennie, C., Call, J., & Tomasello, M. (2009). Ratcheting up the ratchet: on the evolution of cumulative culture. Philosophical Transactions of the Royal Society of London B: Biological Sciences, 364(1528), 2405-2415.

Van Bavel, J. J., & Cunningham, W. A. (2010). A social neuroscience approach to self and social categorisation: A new look at an old issue. European Review of Social Psychology, 21(1), 237-284.

Van Bavel, J. J., Xiao, Y. J. & Hackel, L. M. (2013). Social identity shapes social perception and evaluation: Using neuroimaging to look inside the social brain. In B. Derks, D. Scheepers & N. Ellemers (Eds.) The Neuroscience of Prejudice and Intergroup Relations. (pp. 110-129). Psychology Press.

Watt, P. & K. Stenson (1988). “The street: ‘it’s a bit dodgy around there’: safety, danger, ethnicity and young people’s use of public space,” Cool Places, Geographies of Youth Cultures. London: Routledge, 249-266.

Wilson-Mendenhall, C.D., Barrett, L.F., Simmons, W.K., & Barsalou, L.W. (2011). Grounding emotion in situated conceptualization. Neuropsychologia 49: 1105-1127

Zaki, J., & Mitchell, J. P. (2013). Intuitive prosociality. Current Directions in Psychological Science, 22(6), 466-470.

Zawidzki, T. W. (2013). Mindshaping: A new framework for understanding human social cognition. MIT Press.



[1] I think that I have a relatively simple experiment that would speak in favor of this interpretation of the shooter bias task. If you have an eye tracker, are interested in shooter bias, what to think about cognitive permeation, and want to talk to me, please send me an email.

[2] My argument in this section is focused on the United States, though I appeal to some research that was carried out in the United Kingdom. I believe that similar sorts of effects are likely to show up across many contemporary human populations; but the precise structure of these effects will depend on local phenotypic regularities and local ideological structures.

[3] To get a sense for this fact as a gestalt, see New York Times, 8 July 2015 for images of racial segregation in US cities

[4] Phia Salter and Glenn Adams (2013, 781) open their recent Psychological Compass article on Critical Race Psychology with the following illuminating example of this result. “The 5th edition of the Publication Manual of the American Psychological Association (American Psychological Association, 2001) included a chapter on “Expressing Ideas and Reducing Bias in Language” that provides guidance on the use of non-offensive language when writing about race and ethnicity. For example, the guidelines note that authors should eschew such ethnocentric and offensive terms as Negro or Oriental and instead use more conventional terms like Black/African American or Asian/Asian American. More illuminating was what the passage failed to say. Although the section spanned three pages, it maintained a noteworthy silence regarding appropriate terms for referring to the people who have historically dominated US society–that is, people of European descent whom we will refer to as White Americans”. If it seems odd to you to question the omission of Whiteness in this context, you’ve probably internalized a White racial frame.

[5] I have a hunch about the nature of the attentional bias here, and I have a sense of the kind of experimental data that would support this hypothesis. But I have neither the funding nor the space to run the study; so I must rely on a more coarse-grained sort of data, and withhold my final judgment until I find someone to collaborate with.

[6] Dopaminergic neurons in the basal ganglia generate predictions about the value and distribution of primary rewards, updating these predictions in response to rewards that are better or worse than was expected, given past experience (Schultz, 1998, 2010). Over time, where the value and distribution of rewards remains constant, these predictions converge on relatively accurate representations of particular features of the evaluative landscape. But since the world is complex and unpredictable, additional learning mechanisms must be employed to track fluctuations in the value of rewards, to monitor changes in the probability of gains and losses, and to adjust subjective estimates of risk and uncertainty (Montague et al 2012; Adolphs 2010);

[7] Thanks to Jenni Mueller for repeatedly pushing me on this point.

11 thoughts on “Socialized Attention and Situated Agency”

  1. Thanks, Bryce, for this synoptic and illuminating piece. I just have a couple of questions to draw out themes that interested me.

    One was about spaces being perceived as ‘clean’ and ‘dirty’: you mention this a lot, but always in conjunction with other words like ‘safe’, ‘normal’, ‘orderly’, and their opposites. And while I can clearly see how the evidence you adduce, and the attentional mechanisms you discuss, are connected to the way people perceive danger and normality, I didn’t see any direct discussion of how the cleanliness theme is being processed, or of what evidence exists about how that works. So I was just hoping you could say a bit more about that.

    The other question was about the compositional relation between microaggressions and macroaggressions, and in particular whether macroaggressions should be thought of as actions at all – if so, who or what is their agent? Microaggressions are clearly actions by individuals, for which individuals can be held accountable, but macroaggressions (at least those which are stable patterns of microaggressions) aren’t. Indeed, you might think that part of the point of the notion of microaggressions is to avoid the red herring that each individual who makes an othering comment is personally responsible for the whole history of oppression.

    Of course you might think that macroaggressions then don’t have an agent, and aren’t actions – that they’re more like impersonal happenings, and while we can try to undo or counteract them this is rather like dealing with a natural disaster. But that might also seem to let everyone off the hook a little too easily. So you might instead want to say that the relevant agent is ‘(white) society’, or ‘(some) people, collectively’, or something else larger than an individual. But it’s hard to know how exactly to spell this idea out (e.g. do individual microaggressors act with enough interdependence to count as acting jointly?), so I invite you to opine on it.

  2. Thanks for the questions, Luke! (And thanks to Cameron, John, and Nick for the opportunity to try these positions out in this forum). I think that you’ve hit upon two issues that I definitely need to say more about, and I feel like I’m in a better position to say something about the second issue than I am about the first. But I don’t know that I have answers that will satisfy you in either case—so I’ll keep thinking about both of them.

    The issue you raise about cleanliness is one where I would expect there to be a lot of interpersonal (but culturally situated) variation. It’s also going to be an issue where class-based considerations come to the fore, as do worries about immigration.

    Most of the empirical literature that I know in this general area comes from discussions of immigration; and there are definitely problems in drawing a larger inference from that literature. In his book “Searching for Whitopia”, Rich Benjamin talks about the cleanliness of white spaces—but I’m not quite willing to hang my argument on his claims either. The third place, where I feel like I’m on a bit firmer ground, concerns the discussions of Housing Choice (Section 8) vouchers in the 1990s. It looks like there’s some evidence that people blame changes away from perceived safety/stability/cleanliness as race/class driven. But I haven’t had a chance to dig into that literature carefully to see if that’s right. So, in this paper, I’m deferring to the sociological literature that makes the contrast in the way that I do here, and it’s something that I’ll need to think about more when I expand on this paper.

    Your second question about macroaggressions is super interesting too!
    You say that “Microaggressions are clearly actions by individuals, for which individuals can be held accountable”; I guess I’m less sure about that. A grad student at Georgetown named Nabina Liebow is finishing her dissertation on this topic, and it looks like a really difficult issue to me. I’m not sure that she’s right, but she’s shown that making this claim requires us to back away from many of the assumptions that lie at the heart of discussions of responsibility and agency. But I won’t argue about this here, as you raise another interesting issue that I’d like to explore (I’ll come back briefly to this at the end).

    I think that macroaggressive patterns are more like wars, riots, White flight, and gentrification. They happen over an extended period of time, they rely on the actions of multiple agents interacting at multiple time scales, and they hang together as social structures that generate particular effects. They aren’t carried out by an agent in the sense of a unified ego, but they are the expression of multiple overlapping forces that hang together as a unity because of the causal relations between the component processes. But the boundaries around that structure are vague, and which particular processes are part of the structure will often be unclear (To be honest, that’s roughly my view of individual actions as well, and of individual agents—but that’s another story for another day). We can get closer to a view of what counts as part of the structure by labeling—but even that doesn’t tell us what the underlying causes are.

    Once you’ve named the Thirty Years War, you’ve got relatively clear boundaries for working out what individual behaviors were relevant to how it unfolded; but from inside, it may not be clear whether you’re taking part; and even with full knowledge of all of the happenings in the general vicinity of that war, it may be unclear whether a particular event is part of the war. I guess I would want to say that the Thirty Years War was an action, and that lots of people took part in it, even if it was sometimes unclear whether a particular action was part of the war. That’s what I want to say about macroaggressions—and that’s going to push me deep into the metaphysics, I think (Laurie Paul offered some suggestions about how to go here, using causal modeling; I’m still thinking about that).

    So here’s the take on patterns of structural exclusion, marginalization, and oppression. There are a lot of individual actions that are clearly part of the structure, there are others whose role is kind of unclear, and we only get a good sense of how they hang together by highlighting the pattern as a whole. In the case of racial bias, I think that there will be some people who are just bad actors, a lot of people who are bad actors because of the habitual tendencies that they’ve internalized, and people whose actions may or may not be part of the relevant structures. But my guess is that we’re not very good at determining which role we are playing in these structures. So we probably need a more nuanced theory of responsibility than I have in my pocket (maybe one like Robin Zheng has been developing; and as I mentioned above, Nabina Liebow has a developed a model of responsibility for microaggressions that might do the trick as well, but I’m not sure where I come down on these issues).

    Is any of that helpful?

    1. Thanks Bryce, that’s definitely helpful! I like the analogy to wars, riots, gentrification, and so on.

      What intrigues me about these sorts of phenomena is that while they don’t qualify as joint actions in the usual way philosophers talk about (where each participant has an intention whose very content is that ‘they’ together act), they do show some key features of joint action which seem to make them interpretable and evaluable in a way that a purely impersonal structure wouldn’t be.

      In particular, it seems that in some of these phenomena some sort of mental state (perhaps not a fully conscious one) is i) shared by many individuals, ii) had by many of those individuals because others have it, and iii) contributes to causing and also to rationalising the individual actions they take. So ‘white flight’, for instance, isn’t a product of each white family that decides to move thinking ‘let’s all us white folks leave here en masse’. But it may be a product of the sort of subconscious aversion to racially mixed neighbourhoods that you describe, a state which has content that rationalises leaving, and which each individual has in part because other individuals have it (in the way that you’re describing in the paper).

      In that sense it seems that these collective phenomena may be describable in an intentionalistic sort of way, as ‘motivated’ by collective attitudes and thus as evaluable and judgeable in some of the ways that are appropriate to actions but not to mere happenings.
      Does any of that make sense?

  3. That makes a lot of sense, Luke! One of the things that didn’t make it into this paper, but which I’ll end up addressing in the longer version, is the research on neighborhood preferences, and the way that they interact with the classic segregation models from Schelling. Your social ontology spin on this seems like it would definitely be worth exploring in more detail too! I hadn’t thought about it that way, but that’s definitely a way of linking this up to existing discussions of collective action. As I said above, though, I don’t really know the right way to think about the questions of responsibility here; but I do think that you’re right to think about White flight in this way. For what it’s worth, I also find the reverse process in the case of gentrification to be interesting—and I’ve written a bit on that, which will go into the book that this is going to be part of. That is probably a similar place where this approach to collective action is worth thinking about!

  4. Bryce, I really like this paper. I think it does a lot to illuminate important topics very relevant to issues about bias and discrimination. Thank you!
    I have a comment and a genuine (!) question.
    Comment: your idea in section four that reinforcement learning, at least as usually explained, is conservative is intriguing. And the fairly common saying that the dopamine reaction depends on whether things are going as expected sounds as though it is. But I’m not sure that ‘as expected’ is meant as you are taking it. Thus suppose I am on the search team trying to find Rusty, the red panda who excaped from the National Zoo. As I patrol my assigned area, I may have no expectations about finding him. He could in another area or just strolling down a sidewalk, up a tree, in a backyard, etc. Nonetheless, if I do catch a glimpse of a red panda part, that will be a huge sign that I am closer to the reward than I expected to be. Dopamine burst! So I’ll go in the direction.

    Question: you start out mentioning the predictive coding approach. Both Clark and Friston insist that it can explain cognition and action (action-selection?) with only a small set of operations. And they include the role of affective properties and goals. This view seems pretty problematic to me, and I wonder whether you think all that you describe can be accommodated on a predictive coding/predictive processing model.

  5. Anne,

    Thanks a bunch!

    You’re definitely right that there will be a motivational reaction when you catch a glimpse of Rusty. My guess is that Rusty-shaped objects will be treated (at least in the short run) as salient rewards, and the encounter will trigger an approach motivation. In my paper in the Saul & Brownstein volume, I discuss the ways in which a goal-stimulus can be treated as a reward for a brief period; my guess is that the right story here is a dopamine gating hypothesis. Precisely how that unfolds is a complicated matter, and it depends on the interactions between model-based systems and model-free/Pavlovian systems. I’m not positive that’s the right way to go here, but I think Its consistent with what you say here, and still empirically plausible. Though I haven’t fully internalized the upshot of the paper on subsecond dopamine fluctuations from the Kishida et al paper that I cite above, and that may have bearing on the issue as well.

    I’m really on the fence about the Friston approach. I think it would be cool of it were true, and it would open up a lot of new ways of thinking about mindedness and sociality. But I retain some of my skepticism that I expressed in my reply to the Clark BBS paper (here: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3501287/) and for now, I kind of prefer to stick with the multiple learning algorithm approach because of the grounding in classical learning theory. That said, the algorithms that predictive coders use can capture all of the kinds of responses that I’m interested in, so my guess is that a similar story could be told in terms of active-inferences—and I’ll probably write a version of that argument up when I put this in the book! If you don’t know this paper, this is the one that shows that active-inference models can account for these learning capacities: http://www.sciencedirect.com/science/article/pii/S0301008215000908

  6. Thanks, Bryce. Helpful comments. Let me pursue one and then make a more general comment.

    It’s the second para of your section 4 that seemed to me to rely on the idea of reinforcement involved things going as predicted. Now I’m not sure quite what you are saying.

    My greatest concern with the predictive coding is that it seems out or touch with recent vision science, which, unlike Marr, doesn’t treat getting things right as the end of vision. Vision evolved to aid our meeting our evolutionary needs, and so to be useful. Unlike PC, vision science may take precision as being too costly for the average bear/person.

    Since precisifying can stop at an early stage when needs are met, I’ve been looking at what in PC might catch this. The inferring itself does not seem indexed to needs; needs seem rather to go into the hypothesis group as indicating something like expected input. I am not sure about this. I’d love to know of anything that addresses this sort of problem

  7. Oh, I see that the article you mentioned does look at the question I’m asking. Great. Except I find in it statements such as, ‘Active Inference assumes that organisms act to fulfil prior expectations that encode the (evolutionarily) values of their states (e.g., having access to food).’

    This is a bit hard to understand since so often what’s encoded is a surrogate for an evolutionary value. E.g., we eat because something tastes good. (Many cat owners are wracked with anxiety when their cats suddenly display no interest whatever in their customary food. It does seem they might starve themselves to death. In a real sense, survival-inducing is not a gustatory virtue,) and the challenge for the PC-er is to find the inferential analogues of pleasure’s motivation. It’s early days, but as far as I can see, that might be done in terms of attention.

  8. Anne,

    I don’t know if this will help, but I think that there’s a personal-subpersonal issue that probably needs to be clarified in that section. What I take myself to expect at the person-level may not always be reflected in the operation of the reward circuitry (sometimes it is; especially where reward processing and contentful representations stabilize in ways that generate psychological and behavioral dispositions). Bryce might be not expect to see Rusty walking around in Adams Morgan (though he may feel some undifferentiated sense that he’s on the lookout, or something). But subpersonal expectations will be chugging along in the background processing—with reward circuitry trained on Rusty. That probably requires honing in on Rusty as a reward, probably with a forward-model of Rusty discovery; but my inclination is to think that such a model doesn’t need to bubble up to consciousness to be used.

    Does that help?

    On the precision-coding issue, I guess I’m inclined to see at least Friston’s version of that hypothesis as computationally cheap, and consistent with the possibility of some sorts of sedimented knowledge being grounded in the material structure of the nervous system (I’m not sure that’s sufficient, but Clark argues for a version of it in his reply to Dennett’s Bayesian theory of consciousness). But like you, I want to make a lot more room for biological preparedness in the shaping of our responses to the world!

  9. Bryce, I think we might be at cross purposes on Rusty. I thought you had a neat but possibly wrong idea; namely, that the idea of ‘better than expected’ has a conservative implication. Potentially at least we can only find rewarding what we expect. That seems wrong, but very interesting.

    There have been so many discussions of, e.g., the differences between premises of practical syllogism and those of assertoric ones, I’m fascinated to find people saying, ‘nevermind, just put them all in the ‘expectations box’ and then see everything as adjusting top-down and bottom-up.’ I keep hearing warning signals.

  10. I think I’m somewhere between interesting and wrong on your account, Anne 😉

    In general, I think that we only find those things that we’re attuned to rewarding. But in the short run, we can use modeled possibilities to trick our reward circuits into finding a particual target stimulus rewarding. “For example, a stimulus that typically evokes negative attitudes (such as a rat) can be treated as having a positive value in the context of a currently active goal; but as soon as this goal is inactive, the valence of that stimulus will revert to its default state” (Huebner 2016). Likewise, a stimulus that I don’t typically find super rewarding (such as Rusty) can be treated as having a positive value in the context of a currently active goal; but as soon as this goal is inactive, the valence of that stimulus will revert to its default state.

    I don’t think that gets to your worries about practical and asertoric syllogisms, but hopefully it makes it at least clearer where I’m coming from.

Comments are closed.