Featured Video Play Icon

Explaining Injustice in Speech: Individualistic vs. Structural Explanation

Saray Ayala, San Francisco State University

[PDF of Saray Ayala’s paper]

[Jump to Valerie Soon’s commentary] [Jump to Alex Madva’s commentary] [Jump to Saray Ayala’s replies]


Implicit bias has recently gained much attention in scholarly attempts to understand and explain different forms of social injustice by identifying causally relevant mental states in individual’ minds. Here we question the explanatory power of implicit bias in a particular type of injustice, testimonial injustice, and more generally in what we call speech injustice. Testimonial injustice occurs when the audience deflates a speaker’s credibility due to the speaker’s perceived social identity (Fricker, 2007). We identify two drawbacks of a widely accepted explanation attributing testimonial injustice to prejudices (e.g. implicit bias) in the mind of the hearer, and argue that further understanding of this phenomenon can be gained from a structural explanation that appeals to discursive conventions and interlocutors’ positions in the communicative exchange.

1. Deficiencies of an individualistic explanation

Imagine an all-men conversation about how to properly distribute the weight in the wooden structure of a house. A woman intervenes and proposes a solution. One interlocutor responds: “this is not home décor, this is serious work”. This is an example of testimonial injustice, a subcategory of epistemic injustice (Fricker 2007). Through epistemic injustice, an individual is wronged in their capacity as a knower. In cases of testimonial injustice, an individual is wronged specifically in their capacity as a giver of knowledge, as happens to the woman in the example above.[1] In line with a recent trend in several disciplines focused on implicit bias as the key for the explanation of different forms of social injustice, a widely accepted explanation of testimonial injustice refers to the (explicit or implicit) attitudes of the individuals involved. Miranda Fricker appeals to prejudices in the hearer that “will tend surreptitiously to inflate or deflate the credibility afforded the speaker” (Ibid. p.17), and specifies that “[T]he main type of prejudice (…) that tracks people in this way is prejudice relating to social identity” (Ibid. 27). This biased mind explanation, tracing particular episodes or patterns of injustice to bias in individuals’ minds against speaker’s social identity, is favored by all the psychological and philosophical literature drawing upon empirical research on implicit bias. A tacit assumption in at least some accounts of social injustice that appeal to implicit bias is that in the absence of implicit bias, the injustice would be reduced, or would not even happen.

Our goal in this work is not to question the existence or significance of implicit bias, but its explanatory power.[2] Explanations of injustice invoking implicit biases in individuals’ minds could appropriately account for the phenomenon only in a society where interactions among individuals are not governed by unjust conventions or norms. If, however, the behavior is governed by conventions that program for and are sufficient to produce unjust treatment regardless of idiosyncrasies of individual attitudes, then biased-mind explanations alone can’t adequately account for the contrast between just and unjust societies.

In this paper we outline ways in which structural explanations identifying higher-level social constraints on behavior address an important gap in our understanding of the phenomenon of social injustice. We spell out a structural explanation of testimonial injustice appealing to factors such as the discursive conventions that operate in the communicative exchange, and the positions that speaker and hearer(s) occupy in that communicative framework.[3] We identify two drawbacks of the biased mind explanation. First, it does not specify what happens with the speaker’s words in cases of testimonial injustice; second, it only partially situates the individual in social reality. We discuss each of these drawbacks in detail below, and defend a structural explanation of testimonial injustice and of a broader form of injustice that we call speech injustice, which we define below.

1.1. What happens to the words?

The biased mind explanation does not tell us what happens with the speaker’s words when their credibility is deflated. When a woman tries to assert something but her audience does not grant her the appropriate credibility, what is it that she gets to do with her words? We claim that Rebecca Kukla’s notion of discursive injustice provides an answer (Kukla, 2014).[4] Discursive injustice occurs when a speaker’s speech act is not given the appropriate uptake, in a way that disempowers the speaker. The performative force of the intended speech act gets distorted and the action that the speaker intended to bring about fails, and a different action occurs,[5] usually one that enhances a disadvantage, creates it and/or contributes to its perpetuation.[6] This distortion cannot be attributed to a failure of the speaker to use an appropriate locution.

Let’s consider an example of testimonial injustice and analyze it with the tools of discursive injustice: a non-native speaker of English is attending an academic talk in an English-speaking country. The non-native speaker makes a contribution to the discussion (without any major deviations from standard English grammar and pronunciation that would interfere with comprehensibility of the intervention) but nobody engages with it. Eventually a native speaker raises the same point and this time people engage in a lively discussion.

In the example, a non-native speaker tries to make a proper contribution to a conversation, but fails to do so. There is plenty of research showing that non-native speakers are given less credibility than native speakers (Brown, Giles & Thakerar, 1985; Giles, 1973). Their perceived identity as a non-native speaker deflates their credibility. How does this happen? In general, when we make a contribution to a conversation, we assume that we are proper participants in that conversation, members of the discursive game that is being played. Our audience, however, might not consider us as such. The notion of entreaty helps understand this. Kukla describes entreaties as speech acts that are spoken “in order to be granted status as a speaker with normative standing within a discursive subspace” (Kukla, 2014, p. 9). Entreaties come from outside the discursive game and are masked or direct requests: “Can I play too?”. They do not count as (proper) contributions. If an utterance intended as a proper contribution is instead taken as an entreaty, this prevents the audience from evaluating its content in the same way that a proper contribution would be, for it only counts as a request to be part of the conversation. This suggests a possible way in which the non-native’s credibility might be undermined: their intervention is not given uptake as a proper contribution to the discussion, but is rather treated as an entreaty, and so the lack of audience engagement is better seen as a response not to its content, but to its (perceived) performative force. If not recognized as a proper contribution to the conversation, there is no fact of the matter about whether or not the non-native’s comment is worth engaging. This analysis of testimonial injustice as a distortion of the performative force clearly shows that it wrongs the speaker not only as a giver of knowledge, but also as a doer. When the force of a speech act is distorted in the aforementioned sense, speakers are prevented to do with their words what they intended to. Thus this principle captures a broader kind of injustice, beyond the epistemic kind. Let’s call it speech injustice. Like discursive injustice as defined by Kukla (2004), speech injustice happens when a speaker’s social position trumps the performative force of their speech acts, and they are prevented from doing with their words what they intended to; instead, their words bring about a different action, usually one that enhances a disadvantage, creates it and/or contributes to its perpetuation. Examples of speech injustice include assertions that are given uptake as expressives, or commands that are taken as requests.[7] Speech injustice differs from the phenomenon Kukla identifies in one important aspect: it is governed by conventions. While Kukla is not sympathetic with the idea that the discursive incapacity she is describing is a systematic phenomenon governed by conventions,[8] we emphasize that speech injustice feeds off discursive conventions. Our defense of a structural explanation of speech injustice in Section 2 captures exactly this idea.

1.2. Social situatedness

The second drawback of the biased mind explanation is that it does not appropriately situate the individual in their social reality. We draw on distributed and extended cognition research in the philosophy of mind (Menary, 2010; Sutton, 2010; Gallaguer, 2013) and defend a full-bodied situatedness, according to which the agent is in constant interaction with what is outside of their skin, continuously outsourcing to external (material and abstract) resources, be they technological aids or social norms. Although the biased mind explanation capitalizes on implicit biases and prejudices acquired by an individual from the society, it nevertheless isolates the individual from the social reality during the real time of the episode of injustice. A complete account of hearers’ situatedness in the social reality must consider not only their attitudes and beliefs, even if they are a result of their exposure to social factors, but also the social factors themselves that, at that moment and in that space, are guiding their interpretation of speaker’s words. Appealing exclusively to what is inside the agents’ mind presents them isolated from the social dynamics and the structural forces that shape their behavior in that moment in important ways.

Once we situate the agent in this full-bodied sense, we need to pay attention to how environmental factors are not only facilitating (by providing norms as guides) but also constraining agents’ behavior. When taking part in a conversation or any other social practice, we are part of a social structure. We draw on Sally Haslanger’s notion of social structures as networks of social relations that are themselves constituted by practices (Haslanger, 2015a). These practices locate us in different positions in the structure, that Haslanger calls nodes. From each node, a limited range of possibilities is available (i.e. possible actions to choose from). Someone occupying the node of mother/spouse in a heterosexual relationship in a community with no affordable childcare, and with a male spouse earning a significantly higher salary due to a gender wage gap, has at their disposition a very limited range of possible actions from which to choose when a baby arrives. When the child-bearer decides to quit their job, this decision does not come from an unconstrained decision-making process that straightforwardly reflects their individual’s preferences.[9]

Conversations, qua structured social interactions, situate interlocutors within nodes that afford limited ranges of actions. The affordances of each position within a conversation (e.g., speaker, hearer) are adjusted in virtue of membership in intersecting social categories, more or less salient. For example, a homeless person has a very limited range of actions they can do with their words when addressing a passerby. In an interview to apply for a visa to enter the United States, the status of authority and epistemic privilege of the interviewer, and the burden of proof of a lack of malicious intent lying upon the interviewee together reduce the range of possible speech acts the latter can perform. This range shrinks or expands depending on the combination of different dimensions of social location the person embodies (race, gender, socioeconomic status, level of education, immigration status, country of origin, etc.). For example, a citizen of the United States who tells an immigrant “Speak English, you are in the United States” would be enacting discrimination. If the speaker is however a homeless citizen and the addressee a middle-class immigrant, the speech act has a different force.

The biased mind explanation leaves aside all these structural constrains. In doing so, it paints a very limited picture of the factors governing speech injustice. We next argue that some of those structural factors better explain this kind of injustice.

2. Structural explanation

We mentioned above that the social position of an interlocutor in the broader social structure in which the conversational exchange is situated constrains the range of possible things this person can do with their words. There might be several structural factors playing an explanatory role for each particular social position, or in Haslanger’s terms, for each node, and there are therefore different focuses a structural explanation can take. Here we don’t provide a general analysis of what a structural explanation should look like in order for it to be satisfactory.[10] Our much more modest goal is to point out one type of structural factor that plays an explanatory role in speech injustice. And that is the conventions governing discourse. Conventions can be difficult to individuate and define. For our purposes, it is enough to say that there are conventional norms about what counts as a speech act of a particular type in a given context,[11] together with conventional ways to invoke those norms (e.g. particular words and gestures). In general we easily recognize, by interpreting words and gestures in any given conversational context, what are the conventions that our interlocutor is invoking, and we let those guide our interpretation of (the force of) their speech acts. We humans are in general very sensitive to norms; we are pretty good, from an early age, at detecting them, conforming to them right away, and punishing others for not doing so (Tomasello 2009).

We claim that our ordinary discursive interactions are governed by conventions that systematically distort the speech capacities of people perceived as occupying certain social positions. The existence of these conventions makes speech injustice a feature of well-functioning, although unjust, discursive exchanges. That is, speech injustice is not the result of occasional misinterpretation of some people’s speech acts, but a systematic phenomenon resulting from conventions. Importantly, it is not the result of unskilled listeners who, due to (implicit) biases against speakers’ identity, do not appropriately apply the relevant discursive conventions to that speaker’s intervention. It is rather the unfortunate result of perfectly skilled listeners who are appropriately applying the conventions operative in their communities.

If we are right, people follow unjust conventions and we don’t do anything about it. Could there be any other reason for that, apart from moral laziness? Yes. People are often wrong about the conventions they follow. Let us introduce a distinction that will help unpack this point.

2.1. Operative vs. Manifest conventions

We are sometimes mistaken about the concepts we apply. We might take ourselves to be applying a concept (the manifest concept) that does not actually coincide with the concept we apply in practice (the operative concept, see Haslanger, 2006). For example,

“Consider the term “parent”. It is common, at least in the United States, to address primary school memos to “Parents”, to hold a “Parent Night” or “Parent Breakfast” at certain points during the school year, to have “Parent– Teacher Conferences” to discuss student progress, and so on. However, in practice the term “parent” in these contexts is meant to include the primary caregivers of the student, whether they be biological parents, step-parents, legal guardians, grandparents, aunts, uncles, older siblings, informal substitute parents, etc.” (Haslanger, 2006, p. 99).

Even if the manifest concept used is the notion of immediate progenitor, the operative concept here is the notion of primary caregiver. If one only focuses on the manifest concept, they could be wrong about the concept they think to be applying.

Similarly, our intuitions about which discursive conventions operate in our linguistic communities might be wrong. Sometimes the mistake is easy to notice. On other occasions, however, we might not realize the error. Let’s consider a fictional example: there is a universe with two planets, Earth and Twin Earth, inhabited by people who have antennae. On both planets there is the same manifest convention C about refusal of a sexual approach, let’s say “to reject a sexual approach a person has to perform the standard formula of refusal, which is saying, by assertion, presupposition or implicature, ‘no’”. As it happens, on Twin Earth there is a subpopulation, let’s call them antenna-movers, whose “no”, when intended to refuse a sexual approach, does not get the uptake corresponding to a refusal, unless they utter it while moving their antennae. If they do not move their antennae, their “no” is interpreted in some other way. Thus the operative convention for refusal differs across the planets; on Twin Earth it depends on the perceived antenna-moving status of the speaker. We could define convention C* as a function of convention C plus speakers’ perceived identity. On Twin Earth, the manifest convention for refusal (C) is a generally applicable convention, independent of identity. The operative convention (C*) is, however, constrained by the perceived identity of the speaker. For an antenna-mover on Twin Earth to successfully refuse a sexual approach, the standard formula is not enough, for their speech act won’t be recognized as an act of refusal. They don’t get to deploy the manifest convention C. Importantly, the gap between the manifest and the operative convention is not transparent to inhabitants of Twin Earth. Similarly to this fictional example, it can be the case that our ordinary discursive interactions are governed by operative identity-constrained conventions systematically distorting some speakers’ speech acts.

It is clear now how the phenomenon we identify as speech injustice is different from the one Kukla identifies. She discards the possibility that the distortion of speaker’s performative force is the result of some rules or conventions governing the conversational exchange. Referring to the resulting discursive incapacity, she says “If these effects were sufficiently regular they would become stable conventions in their own right, which could be managed and deployed in the normal way, even if they were politically unfortunate” (Ibid., p. 448). She assumes that the existence of (operative) conventions warrantees interlocutor’s awareness of them, and subsequent capacity to deploy them in a controlled manner. However, once we acknowledge the possibility of operative conventions that might escape interlocutors’ awareness, the existence of these conventions does not guarantee that speakers’ will be able to manage and control their deployment.

Now, how does the existence of identity-constrained conventions help make the case for a structural explanation of speech injustice? One could say that even if we grant the existence of these conventions, they only work through the minds of the individuals, and so an explanation needs to appeal after all to individuals’ mental states. But as we explain below, we don’t need to posit any specific mental state (in particular, implicit bias against speaker’s perceived social identity) to explain the behavior of hearers who conform to (operative) conventions.

2.2. Norm-conforming behavior

When we observe a behavior, the status of the behavior in relation to norms determines the extent to which that behavior can serve as a cue to the agent’s mental states. Prescriptive norms give an agent enough of a reason to behave in accordance to them, and so only when someone behaves in a way that violates a norm, people go on to attribute to the agent some (other) reason to act that way, such as a norm-conflicting mental state (Uttich & Lombrozo, 2010). That is, the existence of a norm that prescribes a particular behavior in a given context can be seen as enough of a reason to exhibit that behavior in that context (no extra reason is required). To borrow an example from Uttich & Lombrozo, an explanation of why a person is wearing a graduation gown at a beach on a hot day (violating the norm) will need to invoke that person’s mental states as reasons for the behavior (perhaps this person likes the gown, or believes the gown makes them look smart, etc.). In contrast, to explain why a person is wearing a graduation gown at a graduation ceremony (adhering to the norm) it is not necessary to postulate a corresponding mental state. An explanation of a norm-conforming behavior does not have to invoke any mental state about the content of the norm – other than a general disposition to conform to it.[12] In fact, Uttich and Lombrozo suggest that is rational to follow such norm-based approach to explanation, for it affords accurate generalization and prediction about related cases.

Knobe (2007) showed that after observing people encouraging kissing between gay partners or interracial sex, subjects attributed intentions to them more often than to those who encouraged kissing between partners of different sexes or sex between partners of the same race. Although the original interpretation of this result appealed to unconscious bias against homosexual kissing and interracial sex, a different explanation seems to be better in accounting for these and other results.[13] Richard Holton proposes that we don’t need to postulate such prejudices in order to explain subjects’ attributions of intentions to the gay-kissing and interracial-sex fans. Subjects’ awareness of social norms condemning those behaviors is enough (Holton, 2010). If gay-kissing and interracial-sex are seen as violating accepted social norms, those who engage in these activities or show sympathy for them will be seen as necessarily having some reason to do it. Those who conform to the norms don’t need any extra reason (i.e. any specific mental state) for their heterosexual kissing and same-race sex preferences. In line with Holton’s interpretation of Knobe (2007)’s results, we claim that we do not need to postulate implicit bias in hearers’ minds in order to explain speech injustice; the existence of (unjust) social norms governing speech interactions is enough to explain this phenomenon.

2.2. What do we gain from identifying structural factors?

We argued that a structural explanation overcomes at least one of the drawbacks we identified in the individualistic explanation, i.e. the lack of situatedness. Moreover, a structural explanation is more stable. There are subtleties in prejudices and implicit bias that surely vary from one individual to the other, which make an explanation of any single episode of speech injustice to be limited to that episode. On the contrary a structural explanation, even though constrained by the cultural context and by eventual variations in discursive conventions, is more stable across individual episodes and individual agents.[14] This stability makes structural explanations better at providing an understanding of speech injustice as a systematic social phenomenon.

Finally, we emphasize the importance of acknowledging structural factors for practical reasons. The presence of implicit bias only makes a difference to the phenomenon of speech injustice when there are no operative conventions systematically distorting speech acts. Effective interventions against speech injustice will have to take into account both individuals’ minds and the conventions that constrain individuals’ behavior.

3. Concluding remarks and directions for future research

We identified two deficiencies of the individualistic, biased mind explanation according to which testimonial injustice results from bias in the hearer’s mind. We proposed that the first deficiency can be addressed with the notion of discursive injustice that appeals to distortion of performative force. This move revealed a kind of injustice that goes beyond the epistemic kind; we called this broader kind speech injustice.

To address the second drawback of insufficient situatedness in the social reality we brought to the scene the structural explanation. This explanation accounts for speech injustice in terms of unjust conventions that constrain interlocutors’ behavior and the positions interlocutors take at any given conversation. We favor structural explanation of speech injustice for its stability and practical relevance, but we emphasize that it is compatible with the existence of implicit bias in hearers’ minds. An important direction of future research will be to consider how structural and individual-level factors – discursive conventions operating in communicative frameworks and individuals’ attitudes – interact in episodes of speech injustice, each party making the most of the other. For example, identity-constrained discursive conventions might become more powerful through individuals’ implicit or explicit identity-prejudices; and more interestingly, individuals’ attitudes might come to existence, and later be reinforced, through individuals conforming to unjust conventions in their discursive practices.


I am grateful to Sally Haslanger and Nadya Vasilyeva for inspiring conversations on structural explanations, and to Andrea Ploder and Ásta Sveinsdóttir for helpful suggestions on a previous draft. I thank attendees at the Hypatia Society meeting at University of California Irvine, the San Francisco Bay Area Workshop on Feminism and Philosophy at San Francisco State University, and the Berkeley Social Ontology Group at University of California Berkeley for comments on earlier versions of this work.


Austin, J. L. 1962. How to do things with words, Oxford: Oxford University Press.

Brown, B. L., Giles, H., & Thakerar, J. N. 1985. Speaker evaluations as a function of speech rate, accent and context. Language and Communication, 5:207–220.

Dretske, F. 1988. Explaining Behavior: Reasons in a World of Causes. Cambridge, MA: MIT Press.

Fricker, M. 2007. Epistemic Injustice: Power and the Ethics of Knowing. Oxford: Oxford University Press.

Gallaguer, S. 2013. The Socially Extended Mind. Cognitive Systems Research 25-26:4-12.

Gilbert, M. 2008. Social Convention Revisited. Topoi 27:5–16.

Giles, H. 1973. Communication effectiveness as a function of accented speech. Speech Monographs 40; 330–331.

Haslanger, S. 2015a. What is a (Social) Structural Explanation? Philosophical Studies 172.

Haslanger, S. 2015b. Social Structure, Narrative, and Explanation. Canadian Journal of Philosophy 45 (1):1-15.

Haslanger, S. 2006. Philosophical Analysis and Social Kinds: What Good are Our Intuitions? Proceedings of the Aristotelian Society, Sup. Vol. 80:89– 118.

Holton, R. 2010. Norms and the Knobe Effect. Analysis 70 (3):417-424.

Keller, E. F. 2010. The Mirage of a Space Between Nature and Nurture. Durham, NC: Durham University Press.

Knobe, J. 2007. Reason Explanation in Folk Psychology. Midwest Studies in Philosophy 31:90-107.

Kukla, R. 2014. Performative Force, Convention, and Discursive Injustice. Hypatia 29 (2)

Langton, R.1993. Speech Acts and Unspeakable Acts, Philosophy & Public Affairs 22: 293-330.

Lewis, D. 1969. Convention: A Philosophical Study. Cambridge, MA: Harvard University Press.

Medina, J. 2013. The Epistemology of Resistance: Gender and Racial Oppression, Epistemic Injustice, and Resistance Imaginations. Oxford: Oxford University Press.

Menary, R. (ed.) 2010. The Extended Mind. Cambridge, MA: MIT Press.

Sutton, J. 2010. Exograms and interdisciplinarity: history, the extended mind, and the civilizing process. In Menary, R. (ed.), The Extended Mind, MA: MIT Press.

Tomasello, M. 2009. Why We Cooperate. Cambridge, MA: MIT Press.

Uttich, K. & Lombrozo, T. 2010. Norms inform mental state ascriptions: a rational explanation for the side-effect effect. Cognition 116:87-100.

Vasilyeva, N. (unpublished manuscript) Situating Structural Explanations


[1] Not all cases of differential attribution of credibility are cases of testimonial injustice, e.g., granting less credibility to a child than to an adult in a conversation about indexicals or law maximizes the communicative value of the conversation without disempowering any of the interlocutors.

[2] Our goal and motivation are similar to Haslanger’s (2015b). She also argues that explanations of social injustice in terms of implicit bias are limited, and defends structural explanations instead. While she addresses social injustice in general, we focus on a specific type of injustice and offer some details of what a structural explanation of this type of injustice would look like.

[3] José Medina (2013) expands on Fricker’s account of testimonial injustice, and points out the need of a less individualistic approach to it. His account nevertheless differs from the one we offer here.

[4] Kukla (2014) does not introduce discursive injustice as the mechanism of testimonial injustice, but as an independent phenomenon.

[5] An extreme case of discursive injustice is Rae Langton’s illocutionary silence, in which the speaker does not get to perform any action (Langton, 1993).

[6] The notion of discursive injustice is based on an important (and widely accepted) assumption according to which audience’s uptake determines what one gets to do with their words (see Austin, 1962).

[7] See Kukla (2014) for several examples.

[8] See Kukla (2004), especially p. 448 and footnote 7.

[9] See Haslanger (2015a) for a detailed analysis of this example, originally in Cudd (2006).

[10] See Vasilyeva (unpublished manuscript) for an analysis of structural explanation that influenced this work.

[11] Very roughly, conventions are regularities we observe, upon which we agree either explicitly or implicitly, and which we deploy either consciously or unconsciously, and through which we organize and coordinate our behavior. Contrary to Lewis’ traditional account (Lewis, 1969), and like many others, we don’t take conventions to require common knowledge nor to be advantageous on all occasions (see e.g. Gilbert, 2008).

[12] We have to distinguish between (merely) following a norm and intentionally following it (see Holton, 2010). The latter might involve adjusting our behavior as to fit the norm, which would probably require beliefs about the content of the norm.

[13] Similar results from many other studies puzzled researchers. See Holton (2010) on why an explanation appealing to norm-conforming behavior accounts better for those results.

[14] See Haslanger (2015a) for more on why a structural explanation is in general more stable than an individualistic explanation.

10 thoughts on “Explaining Injustice in Speech: Individualistic vs. Structural Explanation”

  1. Current research on implicit bias is thought to offer a promising explanation for why social injustice persists. This program assumes the sufficiency of the individualistic biased mind explanation, which claims that injustice is caused by individual mental states. Ayala claims that there are two problems with the individualistic explanation in the case of speech injustice: first, it fails to explain what happens to the speaker’s words, and second, it isolates agents from the social forces that shape their behaviour. Given these problems, Ayala argues that a structural explanation should be explanatorily prior to, and holds more explanatory power than, an individualistic explanation. On the structural explanation, speech injustice is “a systematic phenomenon resulting from conventions” (3) that are appropriate to that society. When someone acts in a way that accords with social norms, there is no need to attribute to them further reasons such as implicit bias. That is just the way things are done, even if we are not aware that we are doing things in this way. My comments will first draw out some additional benefits of Ayala’s structural explanation. I also bring up a few examples where the structural explanation appears insufficient and implicit bias may play an important explanatory role.

    A tacit assumption of appeals to implicit bias is that social injustice is difficult to explain – almost mysterious – in the absence of explicit prejudice. But, as Ayala points out, this approach forgets that the ideal realm in which standard discursive conventions treat all persons as equally credible is not our realm (1). In Ayala’s view, unjust discursive conventions are the appropriate conventions, and those are being applied exactly the way they are supposed to be. This perspective has much to offer both in terms of explanatory power and responsibility for justice, and the first part of my comments will focus on these benefits.

    The first upshot of the structural explanation derives from Ayala’s claim that the structural explanation has more stability than the individualistic explanation. If the structural explanation is more stable, it is also bi-directional. Not only can we explain patterns of injustice by appealing to norms, but we can also appeal to these patterns of injustice in trying to figure out what our operative norms are by reverse-engineering. Patterns of injustice are symptomatic of unjust norms. The individualistic explanation, however, is uni-directional in part because it is less stable. The idiosyncracies of individual bias create a mismatch between consistent patterns and their widely variable causal mental states, and any given explanation is confined to a local context. Thus, we cannot infer backwards from patterns of injustice why some types of speakers consistently get uptake and others do not, regardless of local variations.

    The bi-directional stability of Ayala’s structural explanation also has practical relevance: since it allows us to look at patterns to figure out operative norms, it could help us pinpoint what norms need to be changed. The individualistic explanation constrains social change to the level of the individual and is thus far less efficacious. This is the second upshot of the structural explanation: it is better able to distribute responsibility for injustice and thus motivate social change. It puts the onus on all members of society, particularly those who occupy nodes of relative privilege, to remedy injustice. Appeals to implicit bias sometimes seem to have a tinge of the exculpatory about them – e.g. “I didn’t know I was discounting the words of my Latina colleague, so I’m not really racist.” It is true that it is very difficult, even for the most self-aware people, to know what their implicit biases are. How could one be responsible for changing something that one doesn’t know about? It is also difficult to change these biases even if one is aware of them. By shifting the blame for injustice to isolated individuals, the individualistic explanation contributes to the phenomenon of “racism without racists”, in Eduardo Bonilla-Silva’s (2013) words. What this means is that there are observable patterns of social injustice, but no one believes they are responsible for them (and even if they are implicitly responsible, they don’t know it and therefore can’t change it). There is no such “escape route” for responsibility under the structural explanation. Though we can’t easily access what is going on in our own and other people’s minds, with a little bit of observing and reflecting we can access what is going on out in the world – patterns of injustice and the norms that they imply. If patterns of injustice indicate that conventions are not being uniformly applied across speaker identities, then we are able to reflect on why manifest conventions don’t appear to be applied to certain groups. Granted, this is not that easy. As Ayala acknowledges, the gap between manifest and operative conventions is not always transparent to us (4).

    Let me now offer a few cases involving different sets of norms and intersectional identities in which the individualistic explanation may play an important role. The first case concerns different levels or sets of norms. Distinct sets of norms operate in unison within a diverse society such as the U.S, though there are some broader norms that are applicable to almost all norm-subsets (e.g. a woman’s command will be taken as a request in many subsets). Some norms are more salient in certain communities than in others (e.g. in Asian communities, the words of elders are invested with more authority than in, say, white communities). The subset of norms which applies to the conversation also depends on the hearer’s social identity, and if the norms applicable to the hearer don’t explain her behaviour, the biased mind might indeed have some explanatory role to play. Take a case of speech injustice against a man who does not present as traditionally masculine. If the hearer is an older person who grew up with traditionally masculine norms, we do not need to appeal to implicit bias in order to explain why she discounts his words; that’s just the norm for her. But if the hearer is a metropolitan 20-something who is acculturated to a wider range of acceptably masculine presentations, the norm-based explanation is insufficient to explain why she discounts his words. Thus, in cases where prejudicial behaviour can’t be explained by appealing to norms, the biased mind explanation may step in.

    My second concern regards the sufficiency of the structural explanation for explaining intersectional identities. Ayala cites Haslanger’s (2015) notion of nodes on a social structure to explain what happens with our words. Our social position on a node is the convergence of multiple dimensions of identity, which constrains or expands the range of things we can do with our words (3). The structural explanation seems insufficient in cases of intersectionality, which may be the majority of cases. After all, no one exists along a single dimension of identity. It is doubtful that norms are fine-grained enough to account for complex identities embodied in one person, such as wealthy/queer/immigrant/man of color. When two sets of operative identity-constrained conventions conflict, do we need to appeal to bias to explain why one wins out over another? Imagine the case of an upper-class black woman politician in conversation with a group of middle-class white men. Additionally, imagine that each of her four identity features are equally salient. Some might give her words more uptake because she is clearly upper-class and has power, and it is a discursive convention to treat the words of such people with more credibility. However, others might discount her words because of her race or gender, as it is normal to treat sceptically the words of a black person or a woman. The structural explanation can adequately explain both these different reactions, but in order to account for the difference in reactions between hearers (given the similarity in norms), the biased mind explanation may come into play. More clarification is in order here.

    I end my comments with a practical worry concerning collective change: how are the oppressed to change a structure of unjust discursive conventions? If Ayala’s structural explanation of speech injustice is right, the oppressed appear to be hopelessly stuck. Their words do not do what they intend them to do. Thus, their demands that unjust norms change will be heard not as political demands or as claims about injustice, but as mere complaints that can be dismissed instead of being substantively engaged with, or perhaps worse, as militant threats. Two examples from the 1960s come to mind here: the ways in which political demands from Second Wave Feminists and Black Power radicals were received by mainstream society. Rae Langton (1993, 299) puts this point nicely: “If you are powerful, you sometimes have the ability to silence the speech of the powerless…Let them say whatever they like to whomever they like, but stop that speech from counting as an action.” Those who enact and perpetuate unjust discursive conventions also hold the most power to change these conventions, which puts self-determination out of the hands of the silenced. Perhaps this feature of the structural explanation is what makes speech injustice especially pernicious. Nonetheless, I would like to think that there is some way that the silenced can take discursive justice into their own hands.

  2. Critics of individualistic approaches to discrimination argue that this systemic social problem stems not from the irrationality, selfishness, or prejudice of individual actors, but from underlying structural-institutional forces. Philosophers often voice this criticism of individualism, but less often flesh it out in detail, or offer a substantive, alternative conception of discrimination in anti-individualist terms. Saray Ayala’s rich paper goes beyond hand-wavy gestures to “think more about the social and less about the individual,” and makes the substantive claim that persisting patterns of injustice in discourse depend primarily on shared social conventions and norms rather than biased mental states. Ayala also points toward directions for future research, calling on philosophers and scientists to explore in greater depth the socially situated and embodied aspects of speech injustice.

    Speech injustice occurs, roughly, when a speech act receives inappropriate uptake from others due to the speaker’s social identity, in a way that reinforces broader patterns of social disadvantage. Ayala’s example is a non-native speaker who contributes to the Q&A after a talk, “but nobody engages with it.” Then a native speaker effectively repeats the point, and animated debate ensues. Speech injustice is a descendant of Rae Langton’s (1993) illocutionary silencing, Miranda Fricker’s (2007) testimonial injustice, and Rebecca Kukla’s (2014) discursive injustice.

    Ayala contrasts two potential explanations of speech injustice. One, predominant in recent philosophy and social psychology, appeals to biased minds. Fricker and Kukla trace the injustice in these cases to hearers’ prejudices, which in turn reflect pervasive social attitudes. Kukla (2014, 447) locates the proximate cause of discursive injustice in “subterranean assumptions and habits concerning gender, embodiment, and power that are hard to articulate.” I’ll call this emphasis on bias a B-explanation. Ayala proposes instead that speech injustice is more a matter of conventions and norms. To use Fricker’s example from The Talented Mr. Ripley, when Herbert Greenleaf says, “Marge, there’s female intuition, and then there are facts,” perhaps he is not so much prejudiced against women as he is employing a shared (unjust) convention of discounting women’s testimony or rationality. In Kukla’s terms, there might be a convention of treating women’s testimony as an entreaty to join the conversation, rather than a direct contribution to it. I’ll call these C-explanations.

    Before weighing their relative merits, it’s worth emphasizing a central feature these explanations share. Both B- and C-explanations appeal to unconscious, or at least tacit and not-fully-articulated, phenomena. People might harbor a certain bias, or follow a certain convention, and not be fully aware of it. They might sincerely disavow the bias/convention, and mistakenly deny that their behavior ever reflects it. While implicit biases are often glossed as completely unconscious, evidence suggests that people are aware of them qua “gut feelings” (e.g., Ranganath et al. 2008). Frequently, however, we don’t explicitly notice our biased gut feelings, or appreciate how a given bias is influencing us in a specific context. The same is likely true of tacit conventions. Even if we are not explicitly aware that our behavior follows some shared, rule-governed pattern, we might feel an inchoate sense of discomfort when our behavior falls out of synch with that pattern. A standard example might be norms guiding how far away we stand from interlocutors, which are usually noticed only when we come across, e.g., a close-talker who violates them. Both B- and C-explanations posit these peripherally-conscious phenomena, and thereby allow for conflicts between sincere avowals of egalitarianism and unwittingly discriminatory behavior.

    It’s also worth disambiguating a few different claims that might be at play. One relatively uncontroversial claim:

    Moderate Anti-Individualism: it is not the case that bias explains all cases (or components) of speech injustice.

    Ayala writes, for example: “A complete account of hearers’ situatedness in the social reality must consider not only their attitudes and beliefs, even if they are a result of their exposure to social factors, but also the social factors themselves that, at that moment and in that space, are guiding their interpretation of speaker’s words.” Attitudes are one part of a broader, complex story. I think few defenders of B-explanations would deny this, although the sheer amount of time and words many of us (myself included) have spent focusing exclusively on implicit bias (e.g., as it pertains to individual moral responsibility) rather than on social structures may suggest that we have devoted excessive attention to the individualistic pieces of the puzzle. (We may ourselves be subject to kind of attributional bias or convention, dwelling on those causes of injustice located in individual minds and overlooking situations and structures.)

    Another claim, which makes an important contribution to our understanding of speech injustice and other microinequities:

    Moderate Conventionalism: conventions explain a significant subset of cases (or components) of speech injustice.

    I am persuaded by Ayala that the role of conventions in speech injustice has been underexplored. At times, however, Ayala gestures toward two more controversial claims:

    Strong Anti-Individualism: bias explains no (philosophically or politically interesting?) cases (or components) of speech injustice.

    For example, Ayala writes that B-explanations of injustice “could appropriately account for the phenomenon only in a society where interactions among individuals are not governed by unjust conventions.” Such claims suggest that as long as unjust conventions are in place, appeals to bias make no further contribution to explaining, predicting, or understanding speech injustice (see esp. §2.2 on norm-conforming behavior).

    Another controversial claim:

    Strong Conventionalism: conventions explain all cases (or components) of speech injustice.

    For example, Ayala seems to say without qualification that speech injustice “is governed by conventions.”

    Either of these stronger claims is, I think, a pretty hard sell. Couldn’t there be a) some cases where relatively unbiased minds follow biased conventions, b) some cases where biased minds flout unbiased conventions, and then c) “perfect storm” cases where biased minds follow biased conventions? The conjunction of Moderate Anti-Individualism and Moderate Conventionalism seems on firmer ground, and I agree with Ayala that, “Effective interventions against speech injustice will have to take into account both individuals’ minds and the conventions that constrain individuals’ behavior” (6).

    Now let’s wade into the substantive virtues and vices of B- vs. C-explanations.

    Stability. One central advantage that Ayala cites for C-explanations is stability. Conventions are more stable across times, places, and people than are mere mental states, which can vary dramatically across contexts. B-explanations are unstable in the counterfactual sense that subtle differences in people’s minds and contexts would make a given B-explanation inappropriate. Structural explanations “abstract away from certain features of the individual and allow for ‘inessential perturbations’ in the sequence of events” (Haslanger 2015, citing Garfinkel 1981).

    I wonder, however, whether the relative instability of B-explanations is a virtue rather than a vice. Speech injustices are stable in the sense that they continue despite dramatic declines in overtly discriminatory beliefs, norms, laws, etc. But are they stable in the further sense that they occur regardless of subtle perturbations in minds and contexts? It is not that every comment made by non-native speakers at colloquia is ignored. If the phenomena were so regular and predictable, their existence and relevance might be less controversial.

    One reason for continuing skepticism about implicit bias, microinequities, cumulative (dis)advantage, and the general persistence of subtle and informal discrimination is that effect sizes in empirical investigations are often small (Oswald et al. 2013). Studies on NBA referees (Price and Wolfers 2010) and MLB umpires (Parsons et al. 2011) found patterns of racial bias in calling fouls and strikes, especially in “borderline” cases, but vast quantities of data needed to be analyzed to detect bias. The patterns of discrimination were real, and arguably made a difference to game outcomes, but eluded detection via ordinary spectating. These patterns are stable in that they endure over time, affect many individuals, and crop up in many contexts, but are radically unstable in that one could not predict precisely when they’ll occur or say with confidence that any particular decision was affected by race. The vast majority of decisions were accurate, making it difficult to think that any referees were, consciously or unconsciously, following a racist foul-calling convention. It seems more natural to say that they were slightly biased.

    More relevant to speech injustice, Lilia Cortina and colleagues (2011) found that women, and especially women of color, tended to report experiencing more incivility in the workplace than do men. In many cases, the incivility consisted not in overt harassment but in generic forms of rudeness, such as speaking condescendingly or interrupting a colleague. Cortina also found that women’s disproportionate experiences of incivility made them more likely to intend to quit, reinforcing existing evidence that experiencing a work environment as hostile leads one to quit (e.g., Sims et al. 2005). Again, effect sizes were not overwhelmingly large. Participants reported how frequently, on a scale of 1 (never) to 5 (very often), they experienced each of 10 types of incivility from supervisors or co-workers, including “Paid little attention to your statements or showed little interest in your opinions” and “Doubted your judgment on a matter over which you had responsibility.” A total score of 10 would mean never experiencing any of these incivilities and 50 would mean very often experiencing all 10 types. In one study on the US military, average scores of reported experiences of incivility were:

    Figure 1. Estimated Marginal Means for Gender-by-Race Effect on Incivility
    (Figure 1. Estimated Marginal Means for Gender-by-Race Effect on Incivility)

    White men: 17.95
    African American men: 18.07
    White women: 18.83
    African American women: 19.85

    These differences are both real and really subtle. It’s hard to look at them and think folks are generally following a full-blooded convention, conscious or unconscious, of uncivil treatment toward black women.

    Such microinequities may, then, be highly unstable in that their occurrence is contingent on myriad normatively irrelevant factors and the passing whims of individual minds. For example, Forgas (2011) manipulated participants’ moods before they read a philosophy essay either written by “a middle-aged bearded man in a suit with spectacles” or “a young woman with frizzy hair wearing a t-shirt.” Forgas found that participants in a good mood relied on stereotypes and “gut feelings,” and evaluated the older man’s essay, competence, and likeability much more highly than the young woman’s. However, being in a bad mood induced more vigilant, attentive thinking, and reduced this age/gender bias to statistical insignificance.

    Again, fickle features of minds and contexts made the difference between egalitarian vs. discriminatory treatment. By abstracting away from these perturbations, C-explanations may significantly overpredict the quantity and regularity of discrimination. One way to save C-explanations might be to make them more fine-grained by indexing them to mental states and contexts, e.g., a norm of going with your gut when in a good mood.

    Figure 1: Mood moderates halo effects on the evaluation of an essay. Positive mood increased and negative mood eliminated the halo effect.
    (Figure 1: Mood moderates halo effects on the evaluation of an essay. Positive mood increased and negative mood eliminated the halo effect.)

    Figure 2. The interaction of mood and halo effects on judgments of competence of a writer. Positive mood increases and negative mood eliminates halo effects.
    (Figure 2. The interaction of mood and halo effects on judgments of competence of a writer. Positive mood increases and negative mood eliminates halo effects.)

    In any case, the cross-contextual variability of contemporary discrimination may be closely related to one of its profound harms. The incivility reported by Cortina’s participants is deeply ambiguous: just about everybody interrupts and gets interrupted sometimes, so it is difficult to identify any particular instance of interruption as expressive of bias, as opposed to, say, misplaced enthusiasm. But this ambiguity itself is harmful. Members of stigmatized groups sometimes find ambiguous but potentially biased behavior more unsettling and taxing than outright discrimination (Salvatore and Shelton 2007; Sue et al. 2007). Discrimination might sometimes be easier—to cope with and to combat—if people “knew their enemy” was stable social conventions (or explicit prejudice) rather than inconsistently and ambiguously biased minds.

    Cases more amenable to C-explanations? I believe some cases are better explained in terms of conventions, or at least by psychological or structural phenomena we have not yet found ways to measure. Take Rooth and colleagues’ findings about discriminatory hiring practices in Sweden toward Arab-Muslims and obese individuals, e.g., identical résumés with Swedish names were three times more likely to get callbacks for interviews than résumés with Arab-Muslim names. Implicit measures predicted hiring discrimination over and above explicit measures, but the truth is that neither implicit nor explicit measures predicted much of the real-world discrimination. In one study, a full 58% of employers openly reported a preference for hiring thin people, but this explicit preference predicted none of the real-world discrimination. Maybe conventions/norms can better explain or predict patterns of discrimination that are predicted less well by implicit or explicit bias.

    If so, this raises further questions. For conventions to be more than placeholders standing for “whatever it is, besides bias, that explains these patterns,” it would be great to have clearer ways of measuring or detecting them. Under what conditions are we entitled to posit a convention or norm? One criterion might be that people report following the convention, but this won’t apply for tacit discriminatory conventions. Another criterion regards how regular the behavior is. Another might be if people experience a sense of norm violation when the discrimination does not occur. That is, if a hearer did engage with the non-native speaker’s insightful comment, might others feel irritated that this (by their lights) unperceptive or incomprehensible comment from an outsider was being treated as worthy of discussion? Might members of the audience more vigorously wave their hands to get their own “more pressing” question addressed, or otherwise try to intervene and steer the conversation away from what they take to be a sidetracking non-issue? Perhaps we could build here on Holton (2010) and Uttich and Lombrozo’s (2010) interpretations of Knobe’s side-effect effect. They argue that we are more likely to judge actions to be intentional if they violate a norm than if they conform to it. If people are more likely to think that the hearer who engages the non-native speaker does so intentionally than to think that the hearer who ignores the non-native speaker does so intentionally, then the latter speech injustice might be a full-blooded norm. The parallel prediction (a harder and more tantalizing test case, I suspect) would be a greater likelihood of attributing intention to hearers who ignore rather than engage high-status insiders’ comments. One potential snag: it seems essential to Holton’s account that the norm violation is knowing, that all parties are fully aware of the norm and the violator knowingly disregards it, but such explicit awareness is not involved in Ayala’s paradigm cases.

    Complementarity. As Ayala sometimes emphasizes, these explanations can complement each other rather than compete. One proposal for joining the two is roughly this: the cognitive-affective-motor dispositions that we refer to as “implicit biases” just are the psychological constructs that track these oft-unspoken, oft-unjust discursive norms. Michael Brownstein and I (2012a,b) argued for a claim along these lines. Whereas implicit biases are often depicted as arational associations, we argued that they are highly flexible and sensitive to certain norms. So I sympathize with Ayala’s claim that many cases of speech injustice involve “perfectly skilled listeners who are appropriately applying the conventions operative in their communities.” The skill and norm-sensitivity of implicit biases is often overlooked by philosophers and psychologists, and more research on the socially situated and embodied aspects of these phenomena is clearly necessary.

    Framing the debate. Although we were trying to capture an idea similar to Ayala’s, our response was to propose a revision in conceptualizing implicit biases (a revision in psychological theory), rather than a redirection of attention away from the psychological and toward the social. This leads to my final question. I worry that, in some contexts, framing the issue in terms of individual versus structural explanations can be misleading. A complete explanation should, I believe, make reference to individual as well as social, structural, and institutional phenomena, although which of these aspects is more salient to our local explanatory aims will of course vary. Boo reductive, methodological individualism! Boo reductive social explanations that gloss over psychology altogether! Ayala at times seems to agree.

    Often, the upshot of criticisms of individualism may not be that we should switch from one type of explanation to another (psychological to structural), but that we shift our attention within each of these domains. For example, Ayala’s argument calls inter alia for a shift within psychology: from the study of irrational unconscious biases to the study of skillful norm-following. Understanding speech injustice requires, in part, understanding how individuals learn, follow, sustain, and revise conventions and norms, perhaps in ways that elude introspective awareness. The relevant psychological dispositions are not just stereotypical associations passively imprinted on the mind by “mass media,” but reflect a skillful ability to acquire and seamlessly follow norms. There is a parallel shift in social science: from conceiving discrimination solely in terms of the aggregate of individuals’ preferences and beliefs, to the shared conventions and social structures that constrain behavioral options.


    Agerström, J., & Rooth, D. O. (2011). The role of automatic obesity stereotypes in real hiring discrimination. Journal of Applied Psychology, 96(4), 790.

    Brownstein, M.S., and Madva, A.M. 2012a: Ethical Automaticity. Philosophy of the Social Sciences, doi: 10.1177/0048393111426402.

    Brownstein, M.S., and Madva, A.M. 2012b: The Normativity of Automaticity. Mind & Language 27 (4), 410-434.

    Cortina, L.M., Kabat Farr, D., Leskinen, E., Huerta, M. and Magley, V.J. 2011: Selective incivility as modern discrimination in organizations: Evidence and impact. Journal of Management.

    Forgas, J. (2011). She just doesn’t look like a philosopher. . .? Affective influences on the halo effect in impression formation. European Journal of Social Psychology, 41, 812–817 (2011)

    Fricker, M. 2007. Epistemic Injustice: Power and the Ethics of Knowing. Oxford: Oxford University Press.

    Garfinkel, A. (1981). Forms of explanation: Rethinking the questions in social theory. New Haven: Yale University Press.

    Haslanger, S. 2015a. What is a (Social) Structural Explanation? Philosophical Studies 172.

    Holton, R. 2010. Norms and the Knobe Effect. Analysis 70 (3):417-424.

    Kukla, R. 2014. Performative Force, Convention, and Discursive Injustice. Hypatia 29 (2)

    Langton, R.1993. Speech Acts and Unspeakable Acts, Philosophy & Public Affairs 22: 293-330.

    Oswald, F. L., Mitchell, G., Blanton, H., Jaccard, J., & Tetlock, P. E. (2013). Predicting Ethnic and Racial Discrimination: A Meta-Analysis of IAT Criterion Studies. Journal of Personality and Social Psychology. doi: 10.1037/a0032734

    Parsons, C.A., Sulaeman, J., Yates, M.C., & Hamermesh, D.S. (2011). Strike Three: Discrimination, Incentives, and Evaluation. American Economic Review, 101: 1410–1435.

    Price, J., & Wolfers, J. (2010). Racial Discrimination among NBA Referees. The Quarterly Journal of Economics, 125 (4) 1859-1887.

    Ranganath, K. A., Smith, C. T., and Nosek, B. A. (2008). Distinguishing automatic and controlled components of attitudes from direct and indirect measurement methods. Journal of Experimental Social Psychology,44(2), 386-396.

    Rooth, D. O. (2010). Automatic associations and discrimination in hiring: Real world evidence. Labour Economics, 17(3), 523-534.

    Salvatore, J., and Shelton, J.N. 2007: Cognitive costs to exposure to racial prejudice. Psychological Science, 18, 810-815.

    Sims, C., Drasgow, F., and Fitzgerald, L. 2005: The effects of sexual harassment on turnover in the military: time-dependent modeling. Journal of Applied Psychology 90, 1141-1152.

    Sue, D.W., Capodilupo, C.M., Torino, G.C., Bucceri, J.M., Holder, A.M.B., Nadal, K.L., and Esquilin, M. 2007: Racial microaggressions in everyday life: Implications for clinical practice. American Psychologist 62, 271-286.

    Uttich, K. & Lombrozo, T. 2010. Norms inform mental state ascriptions: a rational explanation for the side-effect effect. Cognition 116:87-100.

  3. I would like to briefly respond to some of the comments Valerie Soon and Alex Madva kindly provided. I am grateful to both of them for taking the time to think through my paper and for the insightful comments and suggestions they gave me.

    Some thoughts about Valerie Soon’s comments:

    1. Soon indicates that an important upshot of a structural explanation of social injustice in general is that “it is better able to distribute responsibility for injustice and thus motivate social change”, for, in contrast to individualistic approaches focused on implicit bias, “[I]t puts the onus on all members of society, particularly those who occupy nodes of relative privilege, to remedy injustice.” Soon warns that often “[A]ppeals to implicit bias sometimes seem to have a tinge of the exculpatory about them.”. It’s true that if discriminatory behavior is explained as a result of implicit bias upon which the agent has no control, attributing responsibility seems jeopardized.

    I sympathize with Soon’s comments on this. An explanation that appeals to conventions necessarily calls upon everyone who is participating in the maintenance of that convention. In Soon’s reflection there are however two different questions that are worth distinguishing. First, one about how difficult is to identify the source of injustice, in order to correctly allocate responsibility; second, one about how palatable a particular attribution of responsibility is (and as a consequence, how good at motivating change). In relation to the former question, Soon says “[T]hough we can’t easily access what is going on in our own and other people’s minds, with a little bit of observing and reflecting we can access what is going on out in the world – patterns of injustice and the norms that they imply.” The idea here is that a structural explanation appealing to unjust conventions would be easier to test than an explanation in terms of (non-conscious) mental states. Although some doubts have been put forward about what is exactly what e.g. the IAT (one of the main tools to measure implicit bias) exactly measures, we seem to be in a good position to say that implicit bias exist and affect many of our interactions. If we agree that some of the discriminatory behavior we observe around us is stable and follows certain patterns, we also seem to be in a good position to say that at least some unjust norms exist and regulate at least some of our behavior. At a first glance, I’m not sure there are good reasons to think that in general it is easier to test a structural explanation, compared to one that appeals to implicit bias.

    About the second question (i.e. how palatable a particular attribution of responsibility is), I agree with Soon that when the injustice is seen as a vice of the norms we follow, it is easier (at least practically) to allocate responsibility in a way that can motivate social change. If instead of an obscure, uncontrollable, out-of-the-stream-of-consciousness entity, the cause of our embarrassing behavior is a collective enterprise we all pursue, there is no easy way out (in terms of causal responsibility) and at the same time an easy way out (in the sense that we don’t have to deal with the fact that deep inside we are really racist, sexist, etc.); as individual contributors we have some responsibility for that behavior. I see at least two advantages from the point of view of motivating social change. First, people would probably react less defensively if portrayed as practitioners of unjust norms that are present in their community, as opposed to a secretly prejudiced individual. Avoiding defensive reactions is definitely a promising strategy to induce social change than the opposite. Second, in general individual, spoon-size contributions to injustice are, given their seemingly negligible effects, difficult to use as motivation for individual counter-action; but if we look at them as part of a regularity a whole collective conforms to, the size of the collective might work now as a powerful motivation.

    2. Soon writes: “in cases where prejudicial behaviour can’t be explained by appealing to norms, the biased mind explanation may step in”. There are two things in relation to this question that come to my mind. First, we need to figure out how to test whether or not there is an unjust convention at play, which might be challenging (e.g. asking people about the norms they are following wouldn’t necessarily give us information about operative conventions, but manifest conventions). Second, when explaining an episode of speech injustice, absence of specific discursive conventions prescribing the specific behavior we want to explain doesn’t disqualify a structural explanation as appropriate or informative. There might be other structural constraints (and in particular, other norms) operating on interlocutors’ positions. Structural constrains might not always work in a straightforward manner. When explaining a particular behavior B, absence of a norm that exactly prescribes B doesn’t mean there are no other structural factors constraining B and in general operating upon the agent’s position.

    3. Soon puts forward an interesting reflection on the complexity of addressing intersectionality. She writes: “The structural explanation seems insufficient in cases of intersectionality, which may be the majority of cases (after all, no one exists along a single dimension of identity). It is doubtful that norms are fine-grained enough to account for complex identities embodied in one person, such as wealthy/queer/immigrant/man of color”. I agree with Soon that we all exists along several dimensions of social identity, and therefore an explanation of types of injustice that operate upon those identities needs to account for that. I do not see, however, how this is a problem exclusive of a structural explanation. We can have the same concerns about implicit/explicit biases: is there a specific (implicit) bias against people who are queer/immigrant/man of color? If so, how is it adjusted when we add some positive dimension like wealthy? Or is it rather the case that we have different biases about different dimensions, and they get combined and adjusted for every context? Which alternative is more psychologically plausible? And which one would be easier to incorporate in an explanation of discriminatory behaviors? If intersectionality is a problem for an explanation of injustice based on social identity, it is a problem for both types of explanation, structural and in terms of biases.

    Some thoughts about Alex Madva’s comments:

    1. Madva distinguishes between different possible views in the individualism/anti-individualism spectrum (moderate anti-individualism, moderate conventionalism, strong anti-individualism, and strong conventionalism), which urges me to clarify my position. I sympathize with what Madva calls “Strong Anti-individualism”, but with an important caveat: only for specific contexts, i.e. communicative frameworks where unjust conventions operate. Let’s call these contexts C-contexts. I would re-state the view as follows:

    [C-contexts] Strong Anti-individualism: in contexts where unjust conventions are in place, a structural explanation of speech injustice is necessary and sufficient.

    First, it is necessary because a full-bodied situatedness in the social reality forces you to consider factors outside individuals’ heads that are both constraining and guiding their communicative interactions. Otherwise you are missing the point. Second, it’s sufficient in such contexts for the reason expressed in §2.2 (when I talk about norm-conforming behavior and refer to Uttich & Lombrozo 2010, Holton 2010, and Knobe 1997). Where there exist norms prescribing certain behaviors, an explanation of that behavior does not need to involve any mental states about the norm (other than a general disposition to conform to it).

    Now what about contexts in which we do not know what’s going on? We can hypothesize that both implicit bias and unjust conventions could be at work. In these other contexts, the most reasonable way to go is to say that a combination of both individualistic and structural considerations has more chances to be explanatorily adequate.

    It is tempting to turn the question about explanation (what is the best explanation?) into a question about causes (what’s the cause of this or that episode of speech injustice, biases or conventions?). I take explanations to be much more than identifying causes. Also, as I state at the beginning of the paper, my goal is not to question the existence implicit bias (and in particular, to question whether implicit bias is at the causal basis of speech injustice), but rather to question its explanatory value for certain types of injustice. Although out of too much excitement about the overarching power of social norms and conventions I do state that speech injustice is more often than not a result of unjust conventions, which constitutes a claim beyond explanation (a statement about how the world actually is, in particular, one that says that more often than not what we find are C- contexts), my goal is explanatory. This is something I definitely need to address in the next version of the paper, for as it is now is certainly confusing. I am grateful to Madva for pointing out this confusion and inviting me to clarify the goal of the paper.

    2. Madva suggests that the stability of structural explanations I point out in the paper, defended by Garfinkel (1981) and also Haslanger (2015), could perhaps be a problem when we face subtle and unstable phenomena, which according to the empirical research he reviews is not uncommon. According to this idea of stability, structural explanations pick up properties that hold across “inessential perturbations” in circunstances (Garfinkel, 1988). But when what we have to explain are subtle and unstable phenomena, Madva considers that perhaps a B-explanation (i.e. that appeals to implicit bias), attentive to the “inessential perturbations” Garfinkel talks about, can actually be the one appropriate. When confronted with the subtle discriminatory behaviors towards black women reported in Cortina et al (2011), it’s hard, Madva says, “to look at them and think folks are generally following a full-blooded convention, conscious or unconscious, of uncivil treatment toward black women”. He continues: “fickle features of minds and contexts made the difference between egalitarian vs. discriminatory treatment. By abstracting away from these perturbations, C-explanations may significantly overpredict the quantity and regularity of discrimination.”

    However, I think the same criticism applies to biases and B-explanations. If the pattern of discrimination is both real and subtle, both an explanation appealing to bias and one appealing to conventions face the same problems. There are two cases that come to my mind: first, one in which there is a lot of noise and the phenomenon we want to identify is unstable. A second case is a complex system where there are many different factors affecting the phenomenon we want to explain. In both cases, an explanation in terms of structural constraints faces the same difficulties that one in terms of mental states. Importantly, appealing to structural constrains does not mean we are turning the system into a deterministic one. Those constraints, whether in the former (noisy) or latter (complex) case, do not determine that certain outcome will happen, but rather make it probable. What a structural explanation reveals is how an object/individual is situated in a network of relationships within a larger whole (structure) and how these relationships modify the probability distribution over possible states of the object/individual, relative to a hypothetical isolated case not embedded in any structure, and/or relative to other positions within the structure, or relative to being a part of a different structure (Haslanger, 2015; Vasilyeva, in prep.).

    A structural explanation of a discriminatory behaviour, whether we are in a noisy or complex system, does not necessarily overpredict the quantity or regularity of the behaviour, for both the noise and the complexity of the system can be accommodated. For instance, we could have a structural explanation of a particular behavior of a particular individual situated in a particular position that tells us that within this structure, this behavior happens 30% of the time. Now, while this is easy to see when talking about structural explanations in general, it is perhaps more difficult to address when talking specifically about explanations that appeal to conventions and norms. It would seem that when norms are in place, the probability that people would follow them is rather high, which then seems to conflict with the cases Madva points out (the noisy and complex cases). However, I don’t think norms and conventions cannot accommodate these perturbations. Having a racist implicit bias doesn’t mean either that you always have an extreme racist behavior in all contexts, nor that you always display racist behaviour along the same dimension or in the same way. Similarly, following an unjust convention doesn’t mean always displaying extreme unjust behavior, or unjust along the same dimension or in the same way. Conventions can be subtle and moderate like bias, and so our adherence to them. And individuals can conform to unjust norms in very different ways, depending on different factors.

    3. Madva indicates that “for conventions to be more than placeholders standing for “whatever it is, besides bias, that explains these patterns,” it would be great to have clearer ways of measuring or detecting them”, and suggests three possible measures we could use to posit when a convention is in place. One criterion he discards is self-reports (that is, people reporting following a convention), for they wouldn’t catch tacit unjust conventions, or to use the vocabulary from the paper that I borrow from Haslanger, operative conventions that are not manifest. Another criterion “regards how regular the behavior is.” This measure seems promising, but in order to link regular behavior with the existence of a convention prescribing it, we will have to be able to discard other causes, like implicit bias, so this criterion wouldn’t be a straightforward measure of conventions, at least it wouldn’t serve to eliminate the alternative possibility of having implicit bias doing the work. The last, more promising criterion Madva considers consists of measuring “if people experience a sense of norm violation when the discrimination does not occur.” Building upon Uttich & Lombrozo (2010)’s empirical results, and Holton (2010)’s interpretation of the Knobe side-effect effect, this third criterion could in turn be applied by measuring how people attribute intentional states to others. If, as Uttich & Lombrozo’s results suggest, we are more likely to judge actions to be intentional when they violate a norm (e.g. wearing a graduation gown on the beach) than if they conform to it (e.g. wearing the gown in a graduation ceremony), patterns of intentional action attribution could be used to reveal the existence of norms. I’m glad Madva points out this possibility, which at a first glance seems to me a promising avenue for an empirical study.

    4. In the last part of his comments, Madva warns that a general framing of the debate as one about individual versus structural explanations can be misleading and not much interesting (a complete explanation of speech and other types of injustice might, at the end of the day, be one that complements both implicit bias and unjust norms). He proposes that a more promising and interesting research project is one about where, within both psychological and structural explanations, we should shift our attention to. He proposes to frame my paper as calling for a shift of attention within psychology “from the study of irrational unconscious biases to the study of skillful norm-following.” I pretty much welcome this suggestion.

  4. Like Soon, I was struck a bit by the apparent hopelessness of Ayala’s account. Where do these unjust conventions come from, and why can’t they be escaped?

    Even though I think that the readings of conversational injustice in terms of convention are in fact insightful and accurate, I’d like to resist a bit by suggesting that Ayala still isn’t quite answering the right question.

    If the category which we are examining is “conversational injustice,” then surely it is right to focus on the highly normative nature of the category. Ayala describes quite a wide variety of phenomena, but what links them is their failure to be just. It is therefore appropriate that we should address this category with a normative question. Not Ayala’s “what is the best way to understand and explain this phenomenon?” but: What is the best way to end these injustices?

    And this is where I think the conventions account falls down, because conversational norms are notoriously hard to change. They are often implicit, unknown to and unnoticed by conversation participants. Direct action on conventions would be extremely hard.

    Direct action on prejudice, though, is possible. And if the prejudices die, then at least the worst of the unjust conversational norms would die with them.

  5. thank you, Phil H, for your comment. It’s true that, like Soon points put in her comments, there is the impression of hopelessness in the diagnosis I propose. However, i think this is just a mistaken impression. These are three reasons why I think so:

    We seem to know more about changing implicit bias: a lot of research is being done on interventions on implicit biases, and so we have some idea of how different biases respond to different types of interventions (we certainly know only a little bit, but we know something). However, we probably know less about how a norm can be changed. This might give us the impression that changing norms is actually more difficult than changing implicit biases. Research on the nature and dynamics of social norms, and in particular on how they can be changed, could eventually gives us good tips about how to change unjust conventions.

    We seem to know less about the nature of implicit bias: we know that social norms change over time, and from that we can infer that they are not tied to any specific “essence” of the society where they are operative, or of the individuals composing that society. However, we don’t seem to know much about the nature of implicit bias, whether they are tied to some “essence”. It’s actually reasonable to entertain the idea that implicit biases express something significant about the individual or individuals endorsing them (this is importantly different from saying that they reflect the individual’s “true or deep self”; you might want to say that implicit attitudes are not reflective of an agent’s deep self, and yet see social norms at a particular time and space as less reflective of individuals living at that time and space, than those individuals’ implicit attitudes).

    Shift of focus: if when designing interventions instead of focusing on the difficulty of changing the origin of injustice (norms/conventions vs. implicit bias), we focus on the question of people’s willingness to accept responsibility and take part in the interventions to end this kind of injustice, then my diagnosis fares better than one that explains speech injustice in terms of implicit bias. It’s easier to accept responsibility for shared conventions (in both a backward and a forward-looking sense), or so I think, than for a set of attitudes you are told you hold which you cannot control. This is my hypothesis, and of course we need to explore it empirically. If my intuition is right, in terms of having all of us see ourselves as part of the problem and also part of the solution, i think my diagnosis is actually not hopeless but promising. Instead of defensive and skeptical reactions, when diagnosed as a result of the norms/conventions we practice we might get willingness to be part of the change.

  6. This is a wonderful paper. It sheds light on something that I have struggled with in law for a while.

    I am a third-year law student at McGill University and am passionate about the criminal law. It is a common occurrence that the credibility of witnesses in court proceedings hinges on their social identity. Many defense lawyers confront witnesses with their criminal records or with their memberships in gangs in order to portray them in a negative light in the jury’s eyes. This is a well-accepted legal tactic, yet most witnesses react defensively to these tactics; they believe them to be biasing.

    Therefore, in light of Ayala’s comment that there should be a shift in psychology, “from the study of irrational unconscious biases to the study of skillful norm-following”, what is there to do when that shift has occurred in a particular domain, say law, but in the wrong direction?

    1. thank you, Diana, for your interest and your comment. The case you describe is very interesting and problematic. I’m not sure, though, that the shift i referred to has occurred in the domain you mention (i.e. legal practice, or court room practices). What i meant with that (and i was actually commenting on Alex Madva’s suggestion) is that we need to pay more attention to the mechanisms of norm-following, both at the psychological and social levels. I didn’t mean (in case that is what you understood) that following norms will alleviate the problems generated by unconscious bias. Is this relevant/helpful for your concern?

  7. A few more thoughts:

    I share Soon’s concerns that intersectional discrimination poses a greater challenge for the convention-based “C-explanations” than the bias-based “B-explanations”. The questions that Ayala raises for B-explanations in this regard are: “is there a specific (implicit) bias against people who are queer/immigrant/man of color? If so, how is it adjusted when we add some positive dimension like wealthy? Or is it rather the case that we have different biases about different dimensions, and they get combined and adjusted for every context?” But it seems to me that these can just be open empirical questions for B-explanations, rather than in-principle problems for them. I believe both are true to some extent: there are many fine-grained biases for specific intersections of identities, and then there will also be complex, messy stories about which aspects of social identity are more salient in which contexts. It seems to me that C-explanations can perfectly well accommodate the fine-grained cases, when there’s a specific convention regarding treatment of a specific intersectional identity. I’m having a harder time seeing how C-explanations will fare when it comes to the messy stories when norms collide. It seems like the way to go would have to be higher-order conventions for deciding which of two competing conventions wins out. Of course, we have lots of higher-order conventions like this in laws, policies, and games, but it’s harder to picture them operating at the kind of tacit level relevant to speech injustice. It’s particularly hard to imagine them being sufficiently shared or regular across different individuals and contexts to qualify as full-blooded conventions rather than biases. (Then again, our unconscious, shared linguistic knowledge may be filled with these higher-order rules…”i before e except after c or…”)

    Regarding Ayala’s revised version of Strong Anti-Individualism:

    “[C-contexts] Strong Anti-individualism: in contexts where unjust conventions are in place, a structural explanation of speech injustice is necessary and sufficient.”

    First, if the claim is that there are specific contexts where C-explanations are more appropriate than B-explanations, I think that, generally speaking, that would be an endorsement of Moderate Conventionalism/Anti-Individualism, because the general view of speech injustice is that some cases admit of C-explanations, some of B-explanations, and some a combination of the two.

    Second, while I’m completely persuaded that appeals to conventions are often necessary for explaining speech injustice, I’m still not convinced that appeals to conventions can be truly sufficient, even if we restrict attention to a narrower range of contexts. Uttich and Lombrozo’s studies are about the psychology of explanation, and show that we are often satisfied with explanations that appeal to norms and that do not appeal explicitly to individuals’ dispositions to follow those norms. But the fact that people tend to find certain explanations satisfying in certain contexts does not yet establish that they are actually good explanations (this is actually a point of emphasis of Haslanger (2015b), following Tilly, on our susceptibility to bogus narratives). And the dispositions to follow norms themselves demand explanation (and understanding these dispositions will be pivotal to combating discrimination). And it might be that implicit biases just are the psychological constructs that follow these unjust norms, in which case they’d creep in even in cases unjust conventions are already in place (although sometimes our local explanatory aims might be such that we are primarily interested in the norm rather than the disposition).

    1. Thank you for your comments. Before trying to articulate a response, a quick thought in relation to Alex Madva’s concern about the sufficiency of conventions to explain speech injustice: it’s true that my appeal to Uttich & Lombrozo (2010) and in general to the psychology of explanation cannot satisfy the normative question of what/how a good explanation is/should be. I discuss this question in a different work, and the general idea is that there is something explanatorily interesting in unreflectively following a norm. The claim here is that people tune up to environmental cues in an unreflective way, that we have the capacity to conform to norms in this unreflective way. Thus, it’s not just a matter of what our folk intuitions are about what explanations sound good, but about the fact that we do conform to norms without necessarily being aware of that (at least once those norms are in place. A different, interesting question is engaging with a new norm, which probably requires deliberation).

      You suggest that “it might be that implicit biases just are the psychological constructs that follow these unjust norms”; we definitely need more research on norm-following, not only psychological research on the dispositions to follow norms but also on how social norms change over time. For now, i think we have enough to suggest that people conform to norms often unreflectively, and that can be the basis for a defense of convention-based C-explanations.

    2. Some thoughts on the necessity of higher-order conventions and the consistency of norms across individuals and contexts. On reflection, I think the concern I raised in my initial comments about intersectional identities (e.g. wealthy/queer/immigrant/man of color) may have been framed in a way that overstates the worry, which was that we don’t seem to have norms that are fine-grained enough to explain the treatment of individuals with complex identities. I want to restrict the scope of this concern, which implicitly assumes that individuals with complex identities are treated relatively uniformly across contexts. This doesn’t seem to be so. People are treated differently in different communities/contexts, and norms don’t compete with each other independently of context. Since norms are relative, as Ayala points out in the graduation gown example, the context will determine which features of an intersectional identity are more salient and the norms that follow. So C-explanations may be sufficient after all. It is stranger to be shabbily dressed at a posh country club than it is to be a person of color (let’s assume that the club is relatively diverse and tolerant), so one might predict that people in that context would be more sensitive to the wealth-related dimension of identity than to race. But I could also imagine the prediction going the other way: since it’s the norm that most everyone at the club is wealthy, that identity feature fades into the background and race gets picked up instead. In either case, we need not appeal to higher-order conventions to decide whether norms about wealth or race win out.

      However, in cases where patterns of injustice concerning an intersectional identity seem to persist across contexts, then I agree with Madva that we’d need to appeal to something like higher-order conventions.

Comments are closed.