Reorienting Implicit Bias Education in Organizations: The MAIBE Checklist

Implicit Bias Can (and Does) Change

It has been argued that organizational implicit bias education is inherently misguided because, once in place, implicit bias is unchangeable. For example, according to Greenwald et al. (2022), “[…] the mental associations that constitute implicit biases are unavoidably acquired from the cultural atmosphere in which one is immersed daily” (p. 9). This argument is problematic for at least two reasons.

First, ample empirical evidence is now available to suggest that implicit bias toward consequential social categories, such as race, age, and sexual orientation is not only temporarily malleable in response to a wide range of interventions but also has the potential to change relatively quickly and noticeably at the societal level (Kurdi & Charlesworth, 2023; Lai et al., 2014). For example, Charlesworth and Banaji (2022) have documented a 65% decrease in anti-gay bias in the United States between 2007 and 2020, whereas Kurdi et al. (2025a) have observed a 20% increase in bias against dark-skinned people internationally between 2009 and 2019. Although demonstrations of enduring change in implicit bias in response to experimental interventions are rare (but see Stier, 2025), such change is clearly possible cognitively (Ferguson et al., 2024). Accordingly, experimentally creating long-term implicit bias change is an ambitious but attainable goal given the current state of knowledge in experimental social psychology.

Second, as discussed in the “Include Measurable Benchmarks” section below, organizational implicit bias education need not be aimed at durably, or even temporarily, modifying employees’ implicit biases. Instead, the goals could range from facilitating individuals’ self-awareness of biased behaviors to improving the retention of minoritized employees and from increasing participants’ familiarity with key insights from experimental social psychology to improving the bottom line by hiring the most qualified — rather than the most stereotypically fitting — person for the job (although for negative downstream consequences of the business case for diversity, see Georgeac & Rattan, 2023). The achievement of each of these goals could be supported by a decrease in implicit bias; however, none of them are predicated on it.

Implicit Bias Measures Have Real-World Analogs

According to a second argument, implicit bias education is inherently misguided because commonly used implicit bias measures, such as the IAT (Greenwald et al., 1998), sequential priming (Fazio et al., 1995), or the AMP (Payne et al., 2005), do not measure unconscious biases, which can have powerful effects in the world. Rather, they measure unintentional biases, which purportedly do not operate outside the lab. For example, according to Gawronski (in press), “[…] there is no real-world counterpart to unintentional biases in the categorization of stimuli on indirect measures” (p. 3). This argument is problematic for at least two reasons.

First, this argument assumes that participants are consciously aware of (that is, they can accurately introspect on) their scores on commonly used implicit bias measures. However, the current evidence supporting introspective awareness of implicit bias is flimsy. Specifically, the well-replicated finding that participants can predict their scores on implicit bias measures above chance (and sometimes even highly accurately; Goedderz et al., 2024) does not imply awareness of the underlying biases. Instead, participants may well be making educated guesses in the same way that they would if they were asked about other people. Supporting this possibility, participants are able to infer others’ implicit biases based on minimal demographic information (Morris & Kurdi, 2023) and struggle to accurately introspect on their own implicit biases newly created in the lab (Kurdi et al., 2025b).

Second, this argument implies that unintentional forms of bias, of which people are aware, do not operate in the real world. This claim is obviously inaccurate, as demonstrated by the three examples cited at the beginning of this paper. For example, the attorney may immediately become aware of the faux pas of asking their lesbian client about her husband instead of her wife; however, once the question has been uttered, the damage to the minoritized individual will already have been done.

Implicit Bias Predicts Behavior; Implicit Bias Change Can Support Behavior Change

According to a third argument, implicit bias education is inherently misguided because commonly used implicit bias measures are poor predictors of, and likely not causally related to, biased behaviors. For example, according to Brauer (2024), “[…] the causal impact of intergroup attitudes on intergroup behaviors is limited” (p. 282). This argument is problematic for at least two reasons.

First, three relevant meta-analyses have established a significant statistical relationship between implicit bias and biased behavior (Greenwald et al., 2009; Kurdi et al., 2019; Oswald et al., 2013). Although the average correlation between implicit bias and behavior is modest on average, it is virtually identical to — if not somewhat stronger than — the relationship between explicit bias and behavior (Kurdi et al., 2019). Moreover, the relationship is reliably stronger in studies that are methodologically rigorous (e.g., those that rely on well-designed indirect measures and measures of intergroup behavior that include both dominant and stigmatized group targets). In addition, further gains in predictive power could be achieved by improving measures of intergroup behavior, recruiting larger samples, and using more appropriate statistical techniques.

Second, although currently no solid evidence suggests that changes in implicit bias are causally responsible for changes in behavior (Forscher et al., 2019), this question is yet to be investigated in methodologically sound, large-scale studies. Moreover, even if implicit bias ultimately turns out to be causally unrelated to behavior, it is difficult to imagine successful debiasing efforts without a change in biased social group attitudes. For example, behavior change may temporarily be maintained by external pressures (such as federal or corporate policies); however, once the external pressure is removed, organizations are likely to return to old practices unless employees are internally motivated to keep relevant changes in place (Plant & Devine, 1998). Indeed, instances of organizations abandoning, and even explicitly renouncing, former inclusion goals have abounded in U.S. higher education and beyond since President Trump issued executive orders targeting diversity, equity, and inclusion programs (The Leadership Conference on Civil and Human Rights, 2025).

Implicit Bias is Compatible with Moral Responsibility for Change

According to a fourth argument, implicit bias education is inherently misguided because relying on the concept of implicit bias absolves perpetrators of discrimination from the moral responsibility for their discriminatory behavior. For example, according to Onyeador et al. (2021), “[…] messaging that attributes discrimination to implicit bias rather than explicit bias reduces accountability and punishment for perpetrators of discrimination” (p. 23).

Although the idea that participants can see perpetrators as less accountable for discrimination emerging from implicit, rather than explicit, bias has received empirical support (Daumeyer et al., 2019; Redford & Ratliff, 2016), this finding does not make implicit bias education inherently worthless or counterproductive. Given that the existence of implicit bias and its importance in group-based discrimination is scientifically well-established, relevant facts should not be withheld from participants of (implicit) bias education programs for fear of how they will respond to this information (Lai et al., 2023).

Rather, the question is how programs can be designed in such a way as to make participants both motivated and empowered to assume accountability for biased behaviors. In fact, moral philosophers have argued that people are just as morally accountable for implicit as they are for explicit biases (Brownstein, 2016). Moreover, feedback about being implicitly biased can motivate change in intergroup behaviors as long as one is willing to accept personal responsibility for their implicit biases (Howell et al., 2024). As such, under the right psychological conditions — which can be facilitated using the appropriate educational tools — learning about implicit bias need not be incompatible with, and can even spur on, egalitarian behaviors.

Implicit Bias is a Feature of Individuals and Environments; Social Change is Predicated on Individuals’ Actions

An emerging and highly influential line of work by Payne and colleagues (2017) has argued that implicit bias should be understood as a feature of environments — such as organizations, cities, and countries — rather than individuals. Supporting this argument, correlations between implicit bias and biased outcomes are often higher at the level of geographic units than at the level of individuals (Calanchini et al., 2022); moreover, implicit bias seems to be more stable over time at the level of geographic units than at the individual level (Vuletich & Payne, 2019). This framework has been used to argue that “[…] we should be focusing more on ways to change the system in which individuals operate rather than trying to change individuals” (Brauer, p. 288). This argument is problematic for at least two reasons.

First, available evidence persuasively demonstrates the value of conceptualizing implicit bias as being situated at the aggregate level, but these demonstrations do not show that individuals are inconsequential. In fact, several pieces of evidence suggest that individuals differ from each other meaningfully and predictably in implicit bias. Among other things, implicit bias varies in expected ways across social group boundaries (for example, between Black and White Americans; Morehouse & Banaji, 2023); implicit bias measures can reveal reliable individual differences if administered repeatedly (Carpenter et al., 2022); and implicit bias can predict individual-level behavior relatively well in methodologically rigorous studies (Kurdi et al., 2019).

Second, even if implicit bias is better understood as situated at the collective rather than the individual level, change in implicit bias and in the consequential outcomes correlated with it is predicated on the (collective) actions of individuals (Madva et al., 2024), especially under societal conditions that are not conducive to facilitating and sustaining such change. Individuals, in turn, can become motivated to effect change only if they are knowledgeable about implicit bias and social group-based inequality more broadly. As mentioned above, top-down organizational change in the absence of individual change in knowledge, values, and intrinsic motivation will likely be fleeting as external circumstances change (Ryan & Deci, 2000).

Include Measurable Benchmarks

The effectiveness of implicit bias education is rarely measured (Lai et al., 2023; Paluck et al., 2021; for notable exceptions, see Hawkins et al., 2023; Onyeador et al., 2025). This practice is understandable if the sole purpose of such education is to ensure that an organization has something to which point in case it is sued for discrimination; indeed, for-profit bias education organizations are ready and willing to provide legal cover for the appropriate remuneration (Dobbin & Kalev, 2016). However, if implicit bias education is to have any purpose beyond offering legal protection, its effectiveness should be measured, for multiple reasons.

First, as with any other type of education, articulating the measurable goals of the activity will make the activity more effective (Anderson & Krathwohl, 2001). Second, if the goal to which the organization aspires is not realistically achievable — such as rooting out implicit bias with a two-hour online course — competent academic psychologists and professionals can inform the organization of this fact and help the organization set more achievable targets. Third, social group-based biases are complex and multifaceted (Schmader et al., 2022), and implicit bias education can serve a multitude of goals, including cognitive, affective, and behavioral ones (Bezrukova et al., 2016). Making educational goals more explicit is the first step toward ensuring that educational activities can be tailored to better fit the priorities of the organization.

Fourth, naturalistic organizational data could provide a valuable new setting for academic psychologists to test predictions of existing theories and to develop new ones. Expanding the scope of research on attitude and stereotype change in this way would be especially helpful given that relevant processes are often studied only in the artificial context of quick online studies involving fictitious targets (Kurdi et al., 2022). Although such research has generated many important insights into basic mechanisms of social learning and memory, its generalizability to real-world settings cannot simply be assumed without empirical testing (Ferguson et al., 2024).

Foreground Epistemic Agency

Educational activities are more likely to succeed if learners are treated as autonomous individuals with epistemic agency rather than empty vessels into which an omniscient educator can pour new information from a position of epistemic and moral superiority (Lai et al., 2023; Reeve, 2013). This insight likely applies all the more strongly to socially sensitive domains, such as social group-based biases, where an educator's (real or perceived) haughtiness (Howell et al., 2024), and even a requirement to attend implicit bias training (Dobbin & Kalev, 2016), can result in backlash effects.

A voluminous and complementary experimental literature now suggests that when participants actively reject information that they encounter, implicit bias change can fail to occur. Specifically, contrary to early theories suggesting that a change in environmental co-occurrences (e.g., exposure to positive Black exemplars or female surgeons) is sufficient to mitigate or even eradicate implicit bias, it has now been demonstrated dozens of times that the way in which participants reason about relevant evidence can considerably impact whether and how implicit biases change (De Houwer et al., 2020). As such, the goals of implicit bias education can easily be thwarted if learners become motivated to reason relevant information away, for example because they feel that organizations or educators do not respect their agency and autonomy.

Integrate Implicit Bias Education into a Larger Toolbox

Implicit bias education will likely be more effective if integrated into a larger toolbox of organizational policies and practices, especially if the goal is to improve outcomes related to the hiring, retention, and belonging of minoritized employees. Such additional measures have been described in detail elsewhere and include (a) mentorship programs (Dobbin & Kalev, 2016); (b) diversity task forces (Dobbin & Kalev, 2016); (c) diversity managers (Dobbin & Kalev, 2016), (d) analyses of existing organizational data to uncover instances of group-based discrimination (Greenwald et al., 2022); (e) preventative measures to interfere with the operation of bias, such as blinding decision-makers to applicants’ demographic identities or removing discretion from decision-making processes (Greenwald et al., 2022; Onyeador et al., 2021; Schmader et al., 2022); and (f) structural changes at the organizational level, including creating opportunities for contact across group boundaries, establishing affinity groups for underrepresented employees, and making organizational messaging more welcoming and inclusive (Onyeador et al., 2021).

The approaches listed above should not be seen as mutually exclusive but are rather likely to be mutually reinforcing as part of a larger toolbox. For example, as argued earlier, organizational policies, such as the creation of affinity groups, will be more likely to endure if they represent individuals’ genuinely held values and intrinsic motivations; these values and motivations, in turn, can be facilitated by appropriately designed organizational efforts. However, more empirical evidence is needed on potential synergies across the tools listed above; notably, synergies might differ depending on the key goals of implicit bias education (e.g., increasing managers’ knowledge levels vs. improving retention of minoritized employees).

Make Implicit Bias Education Broad

The overwhelming majority of implicit bias education programs relies on light-touch, single-shot interventions (Paluck et al., 2021), such as an hourlong online module to be completed by new employees. Whatever the main goal, such efforts are likely to be ineffective. Implicit biases — along with explicit biases, knowledge, and values — are the result of developmental, cognitive, and social processes that unfold over the course of decades rather than minutes or hours. Consequently, hoping to achieve change, and especially lasting change, from one-off activities is psychologically unrealistic (Lai et al., 2016). These considerations are especially pertinent if the goals of implicit bias education also include change in institutional structures, given that institutions often remain calcified even after individual minds and behaviors have changed (Faghih & Samadi, 2024).

As such, if implicit bias education is to have a chance to succeed, it should be broad, in multiple ways. First, it should unfold over the course of months and years, rather than over the course of minutes or hours. Second, in addition to merely providing counterattitudinal information, it should equip participants with the tools to deal with bias-reinforcing information that they are likely to encounter in their daily lives (Devine et al., 2012; Ferguson et al., 2024; Forscher et al., 2017). Third, given the well-known context-sensitivity of learning effects (Gawronski et al., 2018), implicit bias education should become a part of participants’ daily routines, e.g., via an online app downloaded on their phones, rather than remain tied to one particular context. Fourth, if materials follow human-centered design principles and are therefore easy and fun to use (Carey, 2024), participants will be more likely to spontaneously return to them, thus potentially making learning effects more enduring.

Base Implicit Bias Education in Empirical Evidence

Experimental social psychologists have deep disagreements about the potential merits and drawbacks of implicit bias education. Indeed, as demonstrated by the discussions above, no scientific consensus regarding the key questions motivating this review currently exists. However, although scientists’ conclusions can diverge — and sometimes diverge sharply — they all base their conclusions on shared empirical evidence. As such, whenever organizations consider implementing implicit bias education, decision-makers should ensure that any individual or business entrusted with that task follows practices that are informed by, and rooted in, the science of the human mind and behavior. Otherwise, the only benefit of implicit bias education will likely be some degree of protection in lawsuits alleging the violation of anti-discrimination laws (Dobbin & Kalev, 2016), without any further change in organizational outcomes of interest, be it the subjective well-being of minoritized employees, the self-efficacy of those from majority groups in controlling their biases, or the smooth operation of diverse teams.