Hide table of contents

Cross-posted from the Sentience Institute blog. Originally posted November 21, 2018. Some content of the post focuses on the animal rights, farmed animal, and effective animal advocacy communities, but the recommendations are broadly applicable to EA and other organizations and community-building efforts.

The contents of this post are intended to convey research findings only and are not to be regarded as legal advice. Many thanks to Diana Fleischman and Aryenish Birdie for reviewing and providing feedback. Edited by Jacy Reese.



This post summarizes recommendations of strategies to improve equity and inclusion from academic research, government research, and research done in collaboration between consulting firms and companies. I have generally excluded recommendations based on individual lab studies or heavily qualified their conclusions. Most of the findings are either applicable across axes of inequity, were generalized from research on specific axes, or pertain specifically to gender or race (the best-studied axes). The available research is thinner and more reliant on correlational data than one might expect given that the diversity training industry alone is worth $8 billion and more than half the Fortune 500s have diversity programs or officers, but there is sufficient evidence to provide useful guidance to organizations and communities seeking to be equitable and inclusive.

The general takeaway from this research is that organizations will be more effective in efforts to achieve equity if they focus more on implementing inclusionary practices that limit the influence of attitudinal prejudice on behavior than on efforts to directly reduce attitudinal prejudice.

While diversity is well-defined — and in this post, I’ll mostly discuss demographic diversity — there is no consensus in academia or business on precise definitions of equity or inclusion, so I will attempt to lay out definitions that I find both useful and common. “Inclusion” generally has two meanings. It describes the feeling of being included, i.e. being welcomed and treated equitably. Note that this is somewhat distinct from the fact of the matter as to whether a person is being treated equitably, as there can be ambiguity. Inclusion or “inclusionary practice” also describes practices that help us to be equitable, such as heuristics of transparency in decisions where there may be ambiguity about their equitableness, or having an accountability system in place so that any decisions that are made inequitably can be rectified.

Inclusion’s goal of equity is distinct from a goal of equality. Equity is about equality of opportunity, which is not always synonymous with equality of treatment or outcome. Some people have more barriers in the way of their success than others, some of which others have set in place and some of which we ourselves do, and equity is about removing those barriers to even the playing field. Equity is instrumentally important in the same way that being fair and unbiased (which are essentially synonymous with equity) are important: Equity empowers individuals to reach their full potential, and enables teams to benefit from full talent pools. Nondiscrimination, a significant part of ensuring equity, is also a right held by all humans under the United Nations’ International Bill of Human Rights to facilitate human welfare and cooperation.

My own view is that equity and inclusion are important, and generally result in diversity, but that diversity is mostly only important as one of several indicators of equity and inclusion (i.e. if all else were equal, including if the cultural context was one of total equity, I’d be ambivalent between two job candidates who were identical except on a demographic axis[1]). In practice, the evidence herein also suggests that diversity’s instrumental importance is, while generally positive, generally only slightly so, with high variation depending on a group’s goals and the kind of diversity under consideration.

Efforts that merely aim to increase diversity, focusing for instance on the final demographics of a team to the exclusion of other metrics of equity and inclusion, may be counterproductive to the project of equity and inclusion because they may involve shortcuts, tokenizing people from underrepresented groups and counterproductively creating conditions that ultimately fail to recruit and retain excellent employees from underrepresented groups because they won’t help them feel respected, engaged, valued, and like they belong.[2] It is these equitable and inclusive conditions that I think we should strive for.


There is ample evidence of the prevalence of workplace bias, which I won’t belabor in this post,[3] but I would like to take a moment to discuss the “business case” for diversity and to elaborate on my view that equity and inclusion are the more important goals of the DEI trio.

The evidence of correlations between diversity and performance is substantial: An analysis by McKinsey found that “companies in the top quartile for racial and ethnic diversity are 35 percent more likely to have financial returns above their respective national industry medians”; that “companies in the top quartile for gender diversity are 15 percent more likely to have financial returns above their respective national industry medians”; that companies in the bottom quartile both for gender and for ethnicity and race lag in financial performance; that every 10% greater proportion of non-whites on senior-executive teams is associated with 0.8% greater earnings before interest and taxes; and that every 10% greater proportion of women on senior-executive teams is related to 3.5% greater earnings before interest and taxes in the UK.[4] Several recent meta-analyses have shown that gender diversity on a board is slightly correlated with company performance and corporate social responsibility, and that there are small but consistent positive relationships between women in CEO positions or top management teams and long-term company performance on fiscal metrics.[5] Research by human resources consultancy DDI found that companies in the top 20% of financial performance had 37% women leadership, while those in the bottom 20% had only 19%. The Peterson Institute for International Economics found that relative to having no women on the board and in the C-suite, “a 30% female share is associated with a one-percentage-point increase in net margin — which translates to a 15% increase in profitability for a typical firm,” and that while CEO gender was unrelated to performance in their analysis, having more women in the C-suite is. In partnership with Fortune, Great Place to Work also found that the 50 companies ranked best for diversity had 24% higher year-over-year revenue growth than other companies.

Importantly, these are just correlations, and may be explained by diversity improving a company’s performance, but may alternatively be explained by a third variable that makes companies both high-performing and diverse, such as an equitable culture that successfully recruits top talent regardless of demographics; an inclusive culture that fosters belonging for all employees; or the lower number of people from marginalized social groups who currently make it into leadership positions being more impactful than their counterparts from non-marginalized groups because they were subjected to higher standards of competence to access the same positions. In other words, while we can say that diversity — particularly the better-studied cases of gender and racial or ethnic diversity — is correlated to company performance, it’s not clear whether merely increasing diversity — in terms of the demographics of a team — will cause increased performance. Publication bias may also limit this evidence if we expect reports showing negative correlations between diversity and performance to be harder to publish.

There is some evidence that gender diversity does predict revenue independent of employee engagement, and evidence that hiring more women improves the performance of venture capital firms. However, other studies of bio-demographic diversity (as opposed to task-related diversity) show no detectable overall effect on team performance, and mixed effects on innovation (with gains in idea creation apparently getting lost in idea selection and implementation). It’s possible, though, that sufficiently inclusive teams could reap the studied benefits of diversity while mitigating the challenges. And since it seems likely that people from marginalized groups face more day-to-day obstacles, time costs, and stress than those of more privileged groups at the same apparent performance levels, which presumably prevents them from performing as well as they would in a fully equitable world, it’s also possible that the benefits of diversity will increase as communities become more equitable. But these arguments are fairly speculative.

A stronger case for valuing diversity as more than a mere metric of equity is that, relative to corporations, in a social movement, visible demographic diversity may be more important as public relations, figureheads, and interpersonal outreach are all presumably more important to selling an ideology than they are to selling a product, and people might be more readily persuaded and recruited by people they identify with (though they might not be).[6] And of course, the farmed animal movement is a global movement that we want to see scale dramatically, and the US isn’t going to have any single race as the majority within just a few decades, so our movement-building efforts could be negligently limited if we failed to recruit allies from different demographic backgrounds and missed out on massive numbers of potential supporters.

Given all of that, and the reasons stated at the beginning of this post to value equity and inclusion, I think the animal advocacy movement should focus on equity and inclusion much more than diversity (which should still be used as one of several metrics of equity and inclusion, and should also be valued somewhat on its own in the interest of public representation).


Within the animal advocacy movement, to name a few examples of inequity which affect large numbers of advocates and are fairly easy to discern, there is significant room to include and empower people of color; to empower women in leadership; and to ensure the recent swell of interest in and action on sexual harassment results in sustained change. Given the prejudices in society at large, the movement presumably has room to improve on other axes of exclusion too, such as ageism, ableism, and cissexism. My hope is that the recommendations in this document will enable us to improve on all axes of exclusion.

Racial inequity seems worse than gender inequity in the ranks of the movement, as indicated by the large majority of advocates being women while a mere ~10% of staff at farmed animal organizations surveyed by Encompass are people of color. The percentage of people of color in the general population is roughly four times that; rates of vegetarianism are consistent across racial groups; and people of color may even be slightly more opposed to animal farming than white people. So we seem to have significant room to improve racial equity and inclusion.

Rates of people of color and women in leadership appear to have similar rate reductions of around a third, relative to the ranks [7] (while I estimate men have around double the representation in leadership that they do in the ranks), suggesting that leadership has a ceiling of similarly low permeability for both groups.

This potentially higher priority for racial equity may be a mostly immaterial consideration in practice as most of the strategies recommended in this report should reduce inequity generally because they empower people from a multitude of marginalized groups, and there are generally few situations where we have to make tradeoffs between gender and racial equity efforts. When we do, for instance in deciding who to mentor, we can probably usually hit both of these targets at once by prioritizing women of color.

Keep in mind that inequitable behavior within the movement is only part of the explanation for inequities in the movement — though potentially a substantial part — as various external factors may divert people from marginalized groups in earlier parts of the pipeline to the movement.



Organizations should keep an eye on their whole pipeline: attracting, hiring, developing, and retaining talent. Beyond implementing the strategies recommended below, in order to identify team-specific gaps and refine inclusionary efforts large organizations can collect data on the relationships between demographics and, for instance, hiring stages, promotion rates, tenure/turnover, compensation,[8] performance scores, utilization of the professional resources offered by the organization, employee engagement, belonging, and perceptions of the culture. Leadership should commit to addressing any gaps and have some form of accountability in place (e.g. a DEI officer or committee) to ensure they make those efforts. Smaller teams can still try to assess and act on these metrics of equity and feelings of inclusion even though their sample sizes are smaller, conversations may be more suitable than surveys, and they generally have less to gain from such organizational infrastructure work.

Attracting and hiring talent

Research suggests that organizations should:

  • Hire in rounds, evaluating candidates side-by-side, not on a rolling basis.
  • Ensure “requirements” listed in a job posting only list qualifications that are genuinely required, or many qualified women won’t apply even though underqualified men will.
  • Demonstrate a “growth” mindset rather than a “fixed” mindset in job postings. This means using qualifications like “highly motivated” instead of “high performer” or “commitment to improvement” instead of “overachiever.”
  • Try to use gender-neutral or gender-balanced language, or lean a little feminine-typed in job posting word choices, to reduce barriers to application for women. (This does not seem to have a tradeoff with applications from men.) Tools like Textio can help.
  • Make a position’s typical wage range clear, as women are more likely to apply if a wage range, or at least a mention of the room for negotiation, is posted. This may be because they are less confident than men that they will be able to negotiate successfully under ambiguity, which is accurate.
  • Default to “flexible hours” for any roles where that’s possible.
  • In general it’s probably best not to anonymize applications. Field studies generally show no effect on interview selection, and sometimes even show a negative effect (which has also been seen in the lab). Blinding may work for musicians, randomly generated resumes, and identical expressions of interest, but in reality there seem to be subtle cues of an applicant’s background that evaluators may pick up on, and the risk of anonymization backfiring is higher for recruiting groups which are actively interested in DEI. This may be because they are unable to proactively check their biases when blind, or to proactively accommodate disadvantaged candidates at this recruitment stage, or because their staff is already more diverse and people may favor candidates they identify with demographically. There may be circumstances where this is more useful, perhaps such as when recruiters know many applicants personally, and this may or may not also apply to other parts of the hiring process, such as trial projects.
  • If using mental ability tests, use them in tandem[9] with other predictors of job performance to minimize discrimination. A recent meta-analysis suggests that general mental ability tests are the strongest sole predictor of job performance, and that of other predictors tested in combination with it, integrity tests and structured interviews result in the greatest predictive gain, though this research may have significant limitations.[10] Note that per US law, mental ability tests must be nondiscriminatory and a valid predictor of performance for the role. Note, too, that even with those conditions satisfied, their use may still skew applicant pools by, for instance, putting off people of color, who may suspect that the tests will discriminate against them or be used selectively to discriminate against them.
  • In general, looking at context performance in addition to task performance probably makes it less likely that capable candidates from marginalized groups will be inequitably filtered out.

Research has unclear findings regarding:

  • Whether job postings should include an equal employment opportunity statement, which may make the organization look more favorable to candidates from marginalized groups, but may also backfire by indicating potential tokenism. As such, if statements are used it’s probably best if they express that the organization’s commitment to equal employment is out of an interest in assessing applicants equitably, to mitigate that potential concern about a mere interest in superficial diversity.
  • Quotas, for which evidence is mixed.

Other strategies which I have not seen research on, but which follow the general principle of increasing standardization and otherwise limiting selection effects and the influence of biases, include:

  • Publicly post job openings, because private referrals skew the applicant pool towards people like those in organization’s the existing community. Recruiters should generally try to actively source outside candidates, and can use special recruitment programs to recruit people from underrepresented groups.
  • Do your best to start evaluating candidates only after they have seen a job description and applied to the role. Pre-application evaluation can lead recruiters to anchor their views on incomplete, biased, and less relevant information (such as personal fit with the recruiter or ability to think quickly under pressure), instead of allowing each candidate to put their best foot forward.
  • Commit to hiring criteria before posting a job opening or interviewing anyone. Otherwise, recruiters may inadvertently form their criteria inequitably based on stereotypes and personal biases as they encounter applicants.
  • Make these criteria strictly what you are actually looking for, such as “adaptability,” “willingness to update views,” or “creative reasoning” (the latter two of which we’ve used at Sentience Institute) rather than proxies like “start-up experience,” “ivy-league education,” or “previous work at a top marketing firm,” which will unnecessarily restrict the candidate pool, filtering out people who have the skills an organization is looking for but for who did not choose or were inequitably denied access to those specific paths. These resume items can be used as evidence of whether a candidate meets a criteria, but are limiting and insufficient on their own. Relatedly, prefer concrete evidence of whether a candidate meets a criteria, such as by measuring a criteria of “good with donors” with evidence such as “successful experience with major donors” rather than vague and intuitive assessments of whether a candidate is “amiable” (as opposed to too “aggressive” or “abrasive”), which leave ample room for stereotypes and other biases to cloud judgement.
  • Only look for traits that are actually required to perform the role well; otherwise unimportant traits may be prioritized at the cost of important ones. For instance, while basic alignment with an organization’s mission is generally vital, the high-fidelity value alignment that may be shared by the board, founders, or executive leadership (e.g. liberationism at a farmed animal organization that focuses on welfare reforms) is probably not necessary for staff who do not guide the group on big-picture strategy to do their work effectively.
  • Work to avoid exposure to nondiagnostic information about a candidate to limit the introduction of personal biases. My own experience and the experiences of colleagues suggests that recruiters should give little to no value to “cultural fit,” except for the most basic facets such as collaborativeness for a highly collaborative team or adaptability for a rapidly changing work environment. I worry that “cultural fit” often places undue weight — often unintentionally — on unimportant criteria like drinking and socializing habits, shared hobbies, or sense of humor.
  • Make the application process (e.g. steps involved, estimated decision dates) transparent early on to mitigate concerns people from marginalized groups may otherwise anticipate of an unstructured process with too much room for subjectivity and as such discrimination.
  • Score all candidates on a rubric — the same rubric, and the one created beforehand and on which posted qualifications were based. Due to stereotypes and prejudice, it’s possible that people of different demographics tend to be evaluated on different criteria.
  • Relatedly, interview all applicants from marginalized groups who meet the same (or higher) posted qualifications as people from privileged groups who are being interviewed. In other words, set the bar at the same position for every applicant. Since it’s possible that most bias in a talent evaluation pipeline happens in the initial digital stages before candidates are evaluated face-to-face,[11] this may be a particularly important bottleneck.
  • As noted earlier, more underqualified men apply to positions than comparably qualified (or more qualified but still not-quite-qualified) women, which means that loosening criteria when reviewing applications, after that filter, will skew the pool towards applicants who are both men and less qualified compared to if those criteria were loosened in the job posting.
  • Design interview questions that relate as closely and clearly as possible to specific criteria, in order to help evaluators avoid taking less accurate, potentially biased shortcuts. For example, if you need to know how well someone reasons under uncertainty, ask, “What’s a belief you hold weakly, and why do you only hold it weakly?” or use a trial project editing sample research too see if their comments demonstrate how confidently they hold views based on mixed evidence (both of which we’ve done at Sentience Institute).
  • Evaluators should make their initial evaluations separately before coming together to discuss them in order to minimize groupthink and maximize the number of independent perspectives available.

Developing and retaining talent

Research suggests that organizations and managers should:

  • Track engagement and its associations with demographics, and act on their findings. Engagement is correlated with retention,[12] performance, and innovation.
  • To track engagement, use employee survey questions with 5-point likert scales such as:[13]
  • “I feel like I belong at [Organization]”
  • “I am satisfied with how decisions are made at [Organization]”
  • “I feel respected at [Organization]”
  • “At [Organization], there is open and honest two-way communication”
  • “When something bad happens (e.g., when I get critical feedback from my manager, I have a negative social interaction with a peer, etc.), I don’t question whether or not I belong at [Organization]”
  • To improve engagement, work to make employees feel that they have career opportunities, that they have access to the learning and development they need, that the role is as they expected based on the hiring process, that their role contributes to their development, and that their contributions are rewarded and recognized. Inclusive leaders ensure their team members are acknowledged and not micromanaged, give their team the freedom to propose novel ideas and make decisions, give actionable feedback, and listen to feedback from their team. Their team members feel that their voices are heard.
  • Make information about job opportunities and eligibility available to employees, for instance by posting job ladders.
  • Offer learning and development opportunities, such as opportunities for stretch roles (responsibilities beyond an employee’s current role), optional courses, peer-to-peer feedback, fireside chats with industry leaders, programs to help employees develop their life visions and goals, career mapping, company retreats with workshops, and individual budgets to spend as employees choose on learning and development opportunities, after which they can present what they learned to their team.
  • Surveys of employee engagement at Pinterest found that the most inclusive managers: (1) communicate openly about thinking and decision-making with teams and solicit feedback, (2) be transparent about mistakes, (3) solicit input before distributing assignments and give team member latitude in selecting their projects, (4) frequently make time for team and one-on-one socialization, and (5) put in extra time to support struggling team members.
  • Organizations can also work to improve engagement with team lunches with leadership, transparency in executive decision-making, job exchange programs, project retrospectives, and sharing customer/client/user feedback.
  • Make expectations about an employee’s performance clear.
  • Provide feedback to help employees calibrate their confidence to their ability. This is probably of particular benefit to high-performing women on track to work in competitive positions, who may otherwise not have the confidence to compete.
  • In determining pay, use standardized processes that hold evaluators accountable for paying team members fairly.[14]
  • To the extent possible, assign organizational “housework” — voluntary work that is important to the organization but which will not build skills or otherwise accrue points towards promotions — representatively across demographics, in order to avoid such work falling unfairly on the shoulders of women or perhaps other marginalized groups.
  • Establish responsibility for inclusion to avoid collective action failure. The growing trend in the corporate world is for inclusion to be a CEO-level responsibility and not relegated to HR, as recommendations for inclusionary practice are relevant throughout an organization and are not merely a matter of responding to complaints.
  • Establish accountability for inclusion, for instance with transparency policies and diversity managers.
  • Engage managers in the project of inclusion, for instance with mentorship programs and diversity task forces composed of voluntary team members.
  • Do not endeavor to control managers, for instance by forcing them to base their decisions about their team entirely on job tests and performance ratings (though peer evaluation may be good). While it would theoretically be best if managers relied on such objective metrics, forcing them to do so could backfire.
  • Take care with grievance systems, which can backfire by enabling moral licensing and retaliation. This may a hard problem to solve, but perhaps “flexible” complaint systems that use a formal hearing process and informal mediation may be more helpful to employees.
  • Leaders and managers should mentor (i.e. advise) people from underrepresented groups, and should, more helpfully, go beyond that to “sponsor” (i.e. advocate for) them, for instance by offering them connections, recommendations, and other resources beyond training.
  • Diversity trainings generally have no effect on diversity, seem to have a negative effect at least as often as they have a positive one[15] and have no lasting effect on outcomes various researchers have used as proxies for inclusiveness. What specific features and conditions might make a training effective, ineffective, or counterproductive are understudied, but one meta-analysis (which found generally positive effects of trainings and was smaller in scope than the meta-analysis above that found no effect), suggests that they are more effective when they are used in combination with other initiatives, when they target both awareness and skill development, and when they are conducted over a significant period of time.


Research suggests that communities should:

  • Emphasize that members of the community value and, more importantly, are working on equity and inclusion. Social norms are important influencers of behavior within a community, so this will socially incentivize further action.
  • Relatedly, make the successes of organizations doing well on equity and inclusionary practice salient (and not just on the metric of demographic diversity, which may encourage dismissals of “reverse” discrimination, but also on the metrics discussed earlier of engagement, belonging, and mobility across demographics, and in terms of how extensively they are implementing inclusionary practices). We should probably share these successes more than community averages (at least if the average is bad) or the failures of organizations doing poorly, to fuel change by normalizing the good behavior.
  • Actively engage people from overrepresented or more socially privileged groups in the community project of improving equity and inclusion, especially at organizations and in spaces where valuing inclusion is not yet a norm. Unlike people who benefit more and/or more directly from increased equity, they are not likely to be penalized for their advocacy.
  • People from marginalized groups also already bear time and psychological costs from being treated inequitably, so putting this work on their shoulders weighs them down even more unfairly. There is also a more general case to be made for the allyship of people from the privileged group being more effective than advocacy from, or at least solely from, the disadvantaged group.
  • As discussed with regard to empowering team members in organizations, mentor (advise) individuals from underrepresented groups, and more valuably, sponsor (advocate for) them. Even just making connections for people from underrepresented groups is helpful as networking is important to career success.
  • Emphasize the contributions of people from underrepresented groups and otherwise ensure role models and colleagues from underrepresented groups are highly visible.
  • Make visual representations of team members, community members, and target audiences diverse. However, don’t strongly misrepresent a group’s diversity or newcomers may feel uneasy when they realize the group was misrepresented. Showing advocates from underrepresented groups in materials and at events will signal to people who identify with them that they can belong in the movement.
  • Avoid “sex sells” tactics. Studies suggest that when people are prompted to think about women in a sexualized way they underestimate their intelligence and competence. Sexualization may also make men regard women and females as less worthy of moral concern, making them more likely to sexually victimize or behave aggressively towards them. This means these tactics likely contribute a culture of underestimating, undervaluing, and mistreating capable advocates, which inhibits them from reaching their potential.
  • Praise work for the benefit of the community, outside of the scope of one’s designated job responsibilities. Avoid discounting this work in celebrations, hiring, promotion, compensation or other advancement opportunities. Just like salaried positions, this work is a contribution to the community, and women take on more of it (and are expected to) while men are more likely to freeride on women’s contributions. Relatedly, all community members should be encouraged to take on their fair share of community responsibilities.

Other strategies which I have seen little relevant research on, but which are based more on my experience and that of colleagues and which follow the general principle of making people feel they belong, include:

  • In general, avoid too much homogeneity or conformity (in appearance, personal interests, identity affiliation, interpersonal associations, or anything irrelevant to animal advocacy).
  • Especially if you’re in a leadership position, consider “uncovering” yourself if it is sufficiently low personal risk to do so, because it’s less risky than for people lower down the ladder and doing so will signal to others like you that they belong. “Covering” refers to practices which downplay or hide one's identity with marginalized groups, such as by:
  • Altering appearance to conform with the dominant group, e.g. if a black woman loves her natural hair but straightens it to be taken more seriously by white people;
  • Reducing expressions of affiliation with marginalized groups, e.g. if a mother would love to talk about her children but refrains;
  • Avoiding association with other members of one’s group, e.g. if a gay person doesn’t bring their partner to events; or
  • Declining to advocate for one’s group, e.g. if a veteran doesn’t respond to an offensive joke about the military.
  • Some degree of homogeneity even on factors not directly relevant to one’s work may be useful to group cohesion, but it at least seems likely that uncovering one’s identity with marginalized or underrepresented groups will help other members of those groups feel a greater sense of belonging, and that it’s advisable to, for instance, not revolve the community’s social life around frat parties or galas, which appeal to or are accessible to only a very limited subset of the population advocates want involved in the community.
  • Communications of who a “leader” is should move away from their historical stereotypical depictions of masculinity (i.e. their emphasis on dominance, individualism, and heroism) as feminine-typed leadership styles of communality can be as or even more successful leadership strategies. This is not to say dominant leadership styles can have no place — having a diversity of leadership styles seems better than having one leadership style across the movement, whether of a “masculine,” “feminine,” or other type. Relatedly, communications about who our leaders are or can be should include opportunities for people who don’t match the traditional “leader” stereotype of a white man to use a dominant leadership style too.
  • Apologize if you say or do something that is exclusionary and become aware of it, even if it’s something small, such as suggesting a woman smile more, complimenting a religious person’s open-mindedness as an atheist, putting only two genders on a form, or making a joke that suggests potential negative feelings towards people of a marginalized group. While these are individually only minor transgressions, they can accumulate to a substantial burden, and while in some contexts actions like these may not reflect prejudicial views, that’s often unclear, so clearing the air may reduce the effects of that attributional ambiguity.
  • Discuss equity and inclusion issues in ways that don’t further normalize bad behavior or further indicate to people from marginalized groups that they can’t belong here. For instance, while it seems accurate to say, “It’s hard for women to lead in animal advocacy,” or, “The animal advocacy movement has a race issue,” emphasizing these unnecessarily may worsen feelings of exclusion. As such, when possible it’s probably best to focus more on the progress than the problem, though without misleading people into thinking the problem is already solved. For instance, it may be better to say, “Animal advocacy organizations are intent on empowering women in the movement to lead,” or, “The community is working on improving racial inclusion.” Ideally these would be supplemented with concrete examples.


I put this section last because reducing prejudice and bias directly seems more difficult than limiting their influence on our behavior. Bias reduction is also harder to measure, at least in terms of its impact on behavior. Still, there are a few weak findings for the most effective ways to reduce prejudice.

The authors of a 2009 meta-analysis concluded that the following strategies were effective, primarily in reducing implicit prejudices,[16] but also in reducing prejudicial behavior:

  • Cooperative learning
  • Entertainment showing positive intergroup contact and inclusive social norms
  • Peer influence through discussion and dialogue
  • Intergroup contact
  • Value consistency and self-affirmation
  • Cross-cultural training

They also concluded that the following strategies are understudied, ineffective, or have negative effects:

  • Social recategorization
  • Cognitive training
  • Diversity training
  • Multicultural, antibias, and moral education
  • Sensitivity training
  • Conflict resolution

Other research has resulted in the following recommendations:

  • Expose people to favorable ingroup opinions of the outgroup.
  • Intergroup contact can help reduce prejudicial attitudes.
  • Exposure to counterstereotypical examples of individuals from an outgroup (e.g. a woman or black director) can help reduce prejudicial attitudes.
  • Don’t merely raise awareness about bad norms or show stereotypes for educational purposes. As noted in the “Organizational management” section, diversity trainings tend to have no effect on diversity, and while they sometimes succeed, they also sometimes backfire. This seems to be because people are bad at updating on information about biases and seem to consistently think they are an individual exception to normal trends in biases.

The research regarding informing people of their bias or asking people to not be biased is mixed, meaning we should prioritize other strategies, and possibly avoid this entirely at the risk of counterproductive effects.



I’d like to express my pride in the restorative actions that some community members have taken in the face of mistreatment by other community members. It’s important that we create a community that is sustainable and healthy, and it seems to me that we will better achieve that through restorative justice than retributive justice.

“No tolerance” is an important policy for preventing and responding to misconduct, in the sense that every issue will be addressed. But no tolerance doesn’t require a “heavy handed” approach — to the contrary, I think it’s critical for the heaviness of our responses to be commensurate with the severity of an action and to escalate progressively with failures to participate in restorative processes. Small transgressions, for instance, should be “called in”[17] so the person who made the apparent mistake has the opportunity to defend themselves if necessary, or to rectify their mistake and improve. If someone has committed a transgression, we want them to seek understanding, and if they do come to understand, apologize, and demonstrate a credible intention to improve, they should be given the chance to carry forward as a better community member, if possible depending on the severity of their action. To create a culture in which they are incentivized to cooperate like that, any negative consequences resulting from their admission of guilt and cooperation in restoration have to be much less severe and much less likely than the consequences of refusing to cooperate. Otherwise people will have incentive to instead aggressively defend themselves against accusations, which is a lost opportunity for both the healing of the person who was wronged and the betterment of the person in the wrong, and which may result in the disempowerment of everyone involved in ensuing drama.

Of course, restorative justice requires the participation of the accused, and in some cases, when private restoration has been attempted but is failing, it may be necessary to escalate to punishment, “calling out,” or a period of exclusion.

But restorative justice often succeeds, in my own experience and that of others, and it’s what we all want for ourselves when we make mistakes. Because of my own efforts in restorative justice, several people who may have lasting influence in my communities are now more capable allies than they would have been if I had behaved retributively in response to their poor behavior towards me. Instead, they made efforts to understand, make amends, and improve. Private efforts in restoration can take great patience and compassion on the part of the wronged, and that may feel unfair, but the only way we change people is by giving them the opportunity to grow away from the harmful misunderstandings and behaviors that a prejudiced society has taught them, just as we give that opportunity to everyone who used to eat animals or engage in other speciesist and harmful behavior towards nonhumans but pulled their way out of that enculturation — including almost every one of us as farmed animal advocates.

Restoration takes a lot more fortitude and effort in the short run than retribution, but when both the accuser and accused participate it seems to ultimately result in a lot less fighting, stress, time, and harm on all sides, and helps us build a strong and healthy community where people are encouraged to come together and grow rather than a weak and unstable one where we divide and stagnate. So I hope we have the fortitude to work hard for restoration and positive-sum outcomes, even when we’re wounded and rightly frustrated. And I hope we have the courage to accept when our efforts are failing and there will not be a cooperative, healthy resolution, and the resolve to push forward to the least bad outcomes when we fail to reach good ones.


This year, the animal advocacy movement took up the #metoo movement and made notable progress on sexual harassment. Leaders and influencers have been making more public statements of their commitments to DEI. Organizations are seeking guidance from Encompass and other DEI advisors and are developing thorough policies on sexual harassment and nondiscrimination. Women and non-binary advocates in the US organized a productive seminar that took place before the Animal Rights National Conference (ARNC) and are preparing ongoing activities through a Gender Equity in Animal Rights group. DEI was the talk of the town in presentations and conversations at the ARNC this summer, and the conference speaker roster was more demographically representative than in previous years. And one of our largest organizations, Mercy for Animals, is now being led by Leah Garces, who has a track record of caring about and acting on inclusion. And that list is not exhaustive!

I’ve been excited, grateful, and proud to see so much enthusiasm in the animal advocacy community for equity and inclusion, and I’m optimistic about the progress we’ll make, especially with the empowerment of these research findings.

[1] I acknowledge that this is legally what we have to do now anyways, despite the cultural context not being one of total equity.

[2] This is especially true if we seek to include only a token minority of people from underrepresented groups, as there is some evidence that a “critical mass” of team members from socially marginalized groups, including in management positions (maybe 20% in management, at least for women) is necessary to improve team performance relative to a homogenous group, whereas it’s possible that a proportion between zero and that critical mass may harm performance. If this is the case, maybe it owes to “stereotype threat,” anxiety, underconfidence, or a lack of a feeling of belonging for the team members in the small minority, or to the majority’s dismissal of the minority of marginalized team members as tokens. This may only be an effect when a small minority of a team is from a socially marginalized group, whereas groups comprised mostly of people from socially marginalized groups with a minority of people from a socially dominant group (such as a group of mostly women with a small number of men) may fare as well as balanced groups. For instance, men do not seem to demonstrate the increased vigilance, decreased belonging, and decreased desire to participate that women do when their gender is underrepresented in a group, suggesting that there may be no such negative effect of “overshooting” if we end up over-representing people from marginalized groups.

[3] If evidence of inequity in career pipelines has slid past you, here are just a few research findings to start you off. Beyond directly discriminatory decision-making by authorities, people are also alienated from teams and communities when they are mistreated, for instance if they are bullied or sexually harassed, and all of the minor and ambiguous discriminatory interactions people experience can aggregate and amount to significant burdens of stress and feelings of devaluation and exclusion, on top of which cultural stereotypes and the norms and expectations that are cyclically shaped by and shape people push us towards narrow ranges of opportunities and roles that may not be where we would otherwise want to be or can make the most impact.

[4] They also found that racial and ethnic diversity currently have a stronger relationship to financial performance in the US than gender diversity, and that no company is currently in the top quartile on both gender and racial/ethnic diversity.

[5] Studied boards have low percentages of women, so how this extends to or changes with boards closer to or exceeding half women directors is unknown, and there are arguments for why a balanced board could show higher or lower performance. To the extent that prejudice and prejudicially created norms are disempowering women at that stage — e.g. if those who make it through the “glass ceiling” are being dismissed as tokens, aren’t made to feel they belong, and/or only gain entrance to the board in the first place if they meet a higher bar than the men who do — then increasing their numbers will mean less competent men are replaced with more competent women, increasing the board’s performance, perhaps until equilibrium is met around parity. But if pipeline problems create a relatively smaller pool of comparably qualified women than men, then increasing women’s numbers beyond that pool size will result in the reverse.

[6] Note that in the case of anti-smoking PSAs, while demographic similarity (in terms of age, gender, and race) with a smoker character was positively associated with engagement and the perceived effectiveness of a PSA, demographic similarity with a separate non-smoker persuader character was not. This suggests that the persuasive effect of demographic similarity is limited by the extent of deeper context-relevant similarities.

[7] For further clarity, based on the Encompass survey mentioned above, a glance at the National Animal Rights Conference speaker roster, and my memories of many events and knowledge of various groups’ leadership, my estimates are, very roughly, as follows: The ratio of people of color in society compared to the ranks of the movement is 4:1; the ratio of women in society compared to the ranks of the movement is 2:3; the ratio of people of color in the ranks of the movement compared to movement leadership is 3:2; and the ratio of women in the ranks of the movement compared to movement leadership is 3:2. Other groups and intersections are harder to estimate on a glance like this on account of smaller population sizes. For women, the movement's figure of roughly ~75% in the ranks and ~50% in leadership tracks with the general trend in the nonprofit workforce.

[8] Google seems to have done a great job evaluating their pay gaps and immediately addressing them.

[9] Merely hiring the most intelligent individuals may be an ineffective route to developing the most effective team anyways.

[10] The studies it analyzed typically relied on supervisory ratings, which correlate weakly with more objective measures of performance such as output. The study’s methodology also found work sample tests to be poor predictors of job performance, which is suspect as in theory they should very directly measure task performance, and one could choose to measure on-the-job performance by a very similar metric (e.g. a work sample test could be to write an article, and an objective task performance metric on the job could be an evaluation of an article written on the job) just as readily as they might use supervisory ratings. That low ranking of work sample tests and the study’s reliance on supervisory ratings may merely show that supervisors are poor judges of task performance or fail to consider it heavily in their evaluations. The analysis also only looked at general mental ability and combinations of other tests with it, leaving out comparisons between other predictors. Richardson and Norgate have further criticisms of the study’s methodology that limit the weight of its conclusions.

[11] See page 24 of this study. This is very limited evidence and I’m surprised I didn’t find other information about demographic rates within individual hiring pipelines, but this does offer some reason to proactively advance people from marginalized groups at early evaluation stages when it feels like they fall just shy of the bar — because that may just be bias talking.

[12] When engagement is low, employees of a different race than their managers are less likely to want to stay at the company than employees of the same race, but when engagement is high, intention to stay is much higher in both groups, and even higher for the racially diverse dyads than the racially homogenous ones. This both indicates the particular importance of engagement for people from underrepresented groups, and suggests that diversity is an amplifier of engagement (making low engagement lower and high engagement higher).

[13] These five questions correlated best with employee engagement, according to findings based on the use of a Diversity and Inclusion Survey created by Culture Amp, an employee feedback platform, with Paradigm, a DEI consulting firm. Culture Amp has additional question recommendations.

[14] After controlling for compensable factors, organizations may still see discrepancies that may be attributable to gender bias or gender-bias-related challenges such as the double-edged sword women walk when negotiating compared to men. Controlled gender pay gaps within the US are much smaller than the US national pay gap of 78%, which is half explained by gender differences in occupation and largely further explained by other compensable factors such as hours worked, though several percentage points are still unaccounted for by various controlled estimates such as those made by PayScale (0.5-4% depending on the industry, with 1.9% in nonprofits). The national figure indicates inequity in society generally, in industries, and in organizations, but the inequity it points to is largely one of opportunity, caused by direct discrimination throughout pipelines and by the related influences of inequitable social norms and expectations, while it’s only minorly accounted for by direct gender discrimination or otherwise unfair gender-related factors in compensation decisions. Because there still is some unexplained difference in pay for equal work across industries, though, organizations have a responsibility to ensure their compensation is fair. Women also take on more voluntary community work than men, so it’s possible that a controlling variable of “hours worked” fails to capture a significant number of unpaid hours which are still contributed to a company or organization, which would make these apparently-controlled figures inappropriately low estimates of true controlled gaps. This seems particularly relevant in the nonprofit sphere which relies so heavily on voluntary labor. In activist spaces, that voluntary labor may also be significantly comprised of emotional labor for the community and as such may come with a higher stress burden than paid hours.

[15] This may be because they activate stereotypes, make prejudiced behavior appear normal, or make the viewer feel ”woke” for seeing the training or believe that their organization is for showing it, which could enable the viewer to adopt a moral license or relax their care with their behavior.

[16] Implicit prejudice is poorly studied, and popular measurements such as Implicit Awareness Tests (IATs) are poor predictors of behavior. A 2013 review of IATs found “little direct evidence” regarding whether changing implicit prejudice changes discriminatory behavior, that no studies on the correlation between implicit prejudice and discriminatory behavior “reported that implicit prejudice mediated the effect of the manipulation on the behavior,” and that the only study that reported an analysis found no mediation, in addition to which the researchers “found no published paper (successful or not) that tested whether a change in implicit prejudice predicted a later change in behavior.” IATs have low test-retest reliability among other issues, and its founders acknowledged in 2015 that the tests’ failings “render them problematic to use to classify persons as likely to engage in discrimination.” Note that this suggests that changing individuals’ internal beliefs is not a priority for improving equitable behavior in a community (relative to, perhaps, changing their perceptions of community norms, or creating policies and norms and holding people accountable to them).

[17] “Calling in” means addressing a transgression privately, charitably, and probably ideally with “nonviolent” communication, in the interest of decreasing the likelihood of putting the accused on the defensive and increasing the likelihood of a healthy exchange and productive outcome.





More posts like this

Sorted by Click to highlight new comments since:
In general it’s probably best not to anonymize applications. Field studies generally show no effect on interview selection, and sometimes even show a negative effect (which has also been seen in the lab). Blinding may work for musicians, randomly generated resumes, and identical expressions of interest, but in reality there seem to be subtle cues of an applicant’s background that evaluators may pick up on, and the risk of anonymization backfiring is higher for recruiting groups which are actively interested in DEI. This may be because they are unable to proactively check their biases when blind, or to proactively accommodate disadvantaged candidates at this recruitment stage, or because their staff is already more diverse and people may favor candidates they identify with demographically.

I think you are mis-describing these studies. Essentially, they found that when reviewers knew the race and sex of the applicants, they were biased in favour of women and non-whites, and against white males.

I admit I only read two of the studies you linked two, but I think these quotes from them are quite clear that about the conclusions:

We find that participating firms become less likely to interview and hire minority candidates when receiving anonymous resumes.

The public servants reviewing the job applicants engaged in discrimination that favoured female applicants and disadvantaged male candidates

Affirmative action towards the Indigenous female candidate is the largest, being 22.2% more likely to be short listed on average when identified compared to the de-identified condition. On the other hand, the identified Indigenous male CV is 9.4% more likely to be shortlisted on average compared to when it is de-identified. In absolute terms most minority candidates are on average more likely to be shortlisted when named compared to the de-identified condition, but the difference for the Indigenous female candidate is the only one that is statistically significant at the 95% confidence level.

This is also supported by other papers on the subject. For example, you might enjoy reading Williams and Ceci (2015):

The underrepresentation of women in academic science is typically attributed, both in scientific literature and in the media, to sexist hiring. Here we report five hiring experiments in which faculty evaluated hypothetical female and male applicants, using systematically varied profiles disguising identical scholarship, for assistant professorships in biology, engineering, economics, and psychology. Contrary to prevailing assumptions, men and women faculty members from all four fields preferred female applicants 2:1 over identically qualified males with matching lifestyles (single, married, divorced), with the exception of male economists, who showed no gender preference. Comparing different lifestyles revealed that women preferred divorced mothers to married fathers and that men preferred mothers who took parental leaves to mothers who did not. Our findings, supported by real-world academic hiring data, suggest advantages for women launching academic science careers.

This doesn't mean that anonymizing applications is a bad idea - it appears to have successfully reduced unfair bias - rather that the bias was in the opposite direction than the authors expected to find it.

Here is a recent study on the topic that I think is very relevant:

Gender, Race, and Entrepreneurship: A Randomized Field Experiment on Venture Capitalists and Angels (Gornall and Strebulaev)
We sent out 80,000 pitch emails introducing promising but fictitious start-ups to 28,000 venture capitalists and business angels. Each email was sent by a fictitious entrepreneur with a randomly selected gender (male or female) and race (Asian or White). Female entrepreneurs received an 8% higher rate of interested replies than male entrepreneurs pitching identical projects. Asian entrepreneurs received a 6% higher rate than White entrepreneurs. Our results are not consistent with discrimination against females or Asians at the initial contact stage of the investment process.

However, it does seem pretty applicable to EA. The EA community is in many ways similar to the VC community:

  • Similar geographies: the Bay Area, London, New York etc.
  • Similar education backgrounds.
  • Both involve evaluating speculative projects with a lot of uncertainty.

Similarly to the studies discussed above, this finds that people are biased against white men.

(I have some qualms about this type of study, because they involve wasting people's time without their consent, but this doesn't affect the conclusions.)

Blinding may work for musicians


The link you shared does not work, but I assume it was meant to be pointing at the classic study on orchestral interviews from 1997/2000. However, recent re-analysis of the paper (here, here, here, here) shows if anything it supports the opposite conclusion:

This table unambiguously shows that men are doing comparatively better in blind auditions than in non-blind auditions. The -0.022 number is the proportion of women that are successful in the audition process minus the proportion of men that are successful. Thus a larger proportion of men than women are successful in blind auditions, the exact opposite of what is claimed.

The 'fact' that blinded auditions help women overcome bias in non-blinded  auditions came from some dubious pre-replication-crisis analysis, where the authors picked a small subset (often less than three orchestras!) of the data to try to find the effect they are looking for:

The impact of the screen is positive and large in magnitude, but only when there is no semifinal round. Women are about 5 percentage points more likely to be hired than are men in a completely blind audition, although the effect is not statistically significant. The effect is nil, however, when there is a semifinal round, perhaps as a result of the unusual effects of the semifinal round.

... but even with this p-hacking the authors failed to achieve statistical significance. 

So on the whole this suggests that musician interviews is another case where the process was originally biased against men, and blinding helped reduce this bias.

Hi Larks,

Thanks very much for linking that Williams and Ceci article. That was really interesting and quite heartening. I say heartening because I don’t think the bias being shown in that article is unfair. I think the gender of the candidate is a relevant factor in this instance, and in this scenario preferring women when all else is equal will ultimately lead to better outcomes for society.

Those decisions are being made in a context of women being underrepresented in the fields* and I think science is a field where equality in gender representation carries instrumental value. I think this instrumental value comes from provision of new perspectives and minimising blind spots, creating an environment conducive to all people contributing their best, and working towards a stronger applicant pool in the future, one where talented women aren’t discouraged from pursuing careers in these fields. So at this point in time, to me it seems that, all else being equal, being female makes you a more valuable candidate in those fields. This may change in the future if parity in representation is reached; in that case I think it could be unfair and potentially damaging for science and society if there was a persistent bias in favour of females.

To take a different example, I think gender equality is also valuable in school teaching. If I were a school principal and the vast majority of my teaching faculty were female I think I have good reason to prefer a male candidate for a new position if all else was equal in applications.

I think Kelly's recommendations are aimed at someone who has decided that they want to improve diversity in their organisation/field, so it seems fine to be explicit about when tactics are or aren't helpful for this particular aim. She's given some reasons why diversity might be valuable in general but of course the value of diversity will vary depending on the field and context. If you don’t agree that gender equality carries value in science I’d be interested to hear why you hold that view.

*The article notes that in two of the fields (engineering and economics) “women are substantially underrepresented” and in two (biology and psychology) “women are well represented”. Unfortunately I can’t access the cited paper that describes what they mean by “well represented” – some quick googling suggests that women are still under-represented in higher positions in those fields, but feel free to correct that if you have better sources.

The evidence of correlations between diversity and performance is substantial: An analysis by McKinsey found that “companies in the top quartile for racial and ethnic diversity are 35 percent more likely to have financial returns above their respective national industry medians”; that “companies in the top quartile for gender diversity are 15 percent more likely to have financial returns above their respective national industry medians”; that companies in the bottom quartile both for gender and for ethnicity and race lag in financial performance; that every 10% greater proportion of non-whites on senior-executive teams is associated with 0.8% greater earnings before interest and taxes; and that every 10% greater proportion of women on senior-executive teams is related to 3.5% greater earnings before interest and taxes in the UK.[4]

Worth noting that this study failed to replicate:

However, when we revisit McKinsey’s tests using recent data for US S&P 500® firms, we find statistically insignificant relations between McKinsey’s inverse normalized Herfindahl-Hirschman measures of executive racial/ethnic diversity and not only industry-adjusted EBIT margin, but also industry-adjusted sales growth, gross margin, ROA, ROE, and TSR. Our results suggest that despite the imprimatur often given to McKinsey’s (2015, 2018, 2020) studies, caution is warranted in relying on their findings to support the view that US publicly traded firms can deliver improved financial performance if they increase the racial/ethnic diversity of their executives. 

Good post! Regarding casebash's concern about tradeoffs: I think there are clear net benefits to many of these techniques, including matters of basic politeness (e.g. letting people know they are encouraged to bring partners of any gender to events, remembering an "other" option for gender on your forms) and sound business strategy (e.g. only listing actual requirements on your application form, defaulting to flexible hours when that's feasible). If presenting these as "strategies for equity and inclusion" means they're more likely to be adopted, that's a promising development.

Of course, not every organization will benefit from every suggestion, but I like these kinds of "toolbox" posts, which offer a set of options (of varying degrees of implementation complexity) for organizations that want to accomplish something. Almost anyone trying to hire for an EA org is likely to find at least one useful idea here.

(I will note that, while literally every decision a business could make has "tradeoffs", some of these ideas appear especially costly for certain kinds of organizations -- for example, committing to hiring criteria ahead of time might be dangerous if an organization has a lot of work that needs doing and meets someone who is capable of doing A and B, but who applied for a position that does C and D. That said, smaller organizations with more flexible roles and processes can probably work around issues of this nature without much trouble.)

"In general it’s probably best not to anonymize applications. Field studies generally show no effect on interview selection, and sometimes even show a negative effect (which has also been seen in the lab)" - It seems strange to mention this and then not even address the obvious implication that one might draw from this.

The other point is that these practises are analysed as though they don't have tradeoffs, when there almost always is. I suppose discussing this would make this document even longer than it is, but you have listed these as "recommendations" as opposed to "possible approaches".

I'm glad you decided to cross-post this to the Forum after all. :)

"Avoid “sex sells” tactics"—the link there doesn't work.

Curated and popular this week
Relevant opportunities