Unveiling Bias: Assessing the Efficacy of an Intervention in Enhancing Bias Awareness of Pre-Service Teachers

2.1 Attitudes: Stereotypes, Prejudices, and Bias

In social psychology, the terms stereotypes, prejudices, and bias describe related but distinct aspects of how people perceive and evaluate others based on social group membership. From the perspective of the three-component model of attitudes (Rosenberg & Hovland, 1960), stereotypes represent the cognitive component (beliefs), prejudices the affective component (feelings), and bias often reflects the behavioral component (actions or tendencies). Together, they contribute to anoverall evaluative attitude toward social groups, though not all components are inherently evaluative. Understanding the differences between these concepts is crucial for clarifying their roles in social cognition and behavior. Stereotypes are generalized cognitive beliefs or assumptions about the traits, behaviors, or characteristics of members of particular social groups (e.g., Bourne & Ekstrand, 1992; Fiske & Tablante, 2015). They serve as mental shortcuts to simplify social information processing by categorizing individuals, thereby reducing complexity and helping to anticipate social interactions (Lippmann, 1922; Strasser, 2012). However, stereotypes often ignore individual variability and can become rigid and resistant to change (Strasser, 2012). While they may contain an evaluative component, stereotypes do not necessarily involve affective responses. Prejudices are evaluative attitudes, often negative, toward individuals based on their group membership (Allport et al., 1954; Fiske & Tablante, 2015). While prejudices reflect emotional responses toward a group, stereotypes are generalized beliefs that may influence perceptions and judgments without necessarily involving personal feelings or direct experience (Allport et al., 1954; Fiske & Tablante, 2015). The concept of bias is more complex and has been interpreted in different ways. According to Gawronski et al. (2022), bias should be understood primarily as a behavioral component – that is, a reaction or response triggered by social cues. This means that bias is not something one “has” as a static mental representation (unlike stereotypes or prejudices, which are cognitive and affective components), but rather a manifestation in behavior influenced by these underlying processes. This behavioral perspective highlights that bias reflects how stereotypes and prejudices may translate into actions or judgments, sometimes without conscious intention. Consequently, biases influence social interactions and decisions in subtle yet significant ways (De Houwer, 2019; Schwarz, 2015; Vuletich & Payne, 2019). It is important to note that stereotypes, prejudices, and biases can manifest both explicitly (consciously) and implicitly (unconsciously; De Houwer, 2019; Schwarz, 2015; Vuletich & Payne, 2019). Research indicates that while explicit manifestations have declined in recent decades, implicit manifestations persist more stubbornly, continuing to affect across many social domains (Charlesworth et al., 2023; Charlesworth & Banaji, 2022). Finally, bias awareness – the capacity to recognize and critically reflect on one’s own behavioral biases based on the underlying stereotypes and prejudices – is a critical step in reducing their negative impact on social judgments and interactions (Perry et al., 2015). In summary, stereotypes and prejudices serve as cognitive and affective foundations, while bias represents the behavioral expression of these underlying processes. Understanding these distinctions is essential for developing targeted interventions to reduce discrimination and promote equitable social outcomes.

2.2 Awareness as a Mechanism to Reduce Biases Based on Stereotypes and Prejudices

Numerous studies have sought to reduce biases in judgment and to counteract the disadvantages these biases produce for marginalized groups (Lai et al., 2016). However, a definitive solution remains elusive. A common approach involves interventions that aim to dismantle stereotypes and reduce their influence on perception and judgment. For example, brief science-based online diversity training in workplace settings has been shown to lead to changes in attitudes and, in some subgroups, modest behavioral shifts, but its effects are often limited and insufficient as stand-alone solutions for promoting equality (Chang et al., 2019). Many of these interventions focus on increasing awareness, either by emphasizing contextual factors, providing guidelines for intergroup interactions (Hill & Augoustinos, 2001; Stephan & Stephan, 2001), or facilitating reflection processes (Brookfield, 2017). Yet, because stereotypes and prejudices are often activated automatically and unconsciously (Devine, 1989), increasing deliberate awareness of these internalized attitudes, i.e., bias awareness, is essential for mitigating their potential influence on judgments and behavior In this context, awareness of personal bias – meaning the individual recognition of one’s own stereotypes and prejudices – plays a particularly crucial role. Research has shown that bias awareness, defined as the sensitivity to and recognition of one’s own stereotypes and prejudices, varies considerably among individuals (Perry et al., 2019; Perry et al., 2015). Individuals with greater bias awareness are more likely to view feedback about their biases as credible and to take corrective actions (Perry et al., 2015). This form of personal insight is therefore a key factor in long-term behavioral change. One promising strategy is the “habit-breaking” approach, which treats implicit bias as a habit that can be disrupted through a combination of awareness, concern, and intentional behavioral strategies. For example, Devine et al. (2012) showed in a 12-week longitudinal study that participants who underwent such an intervention demonstrated significant and long-term reductions in implicit racial bias, particularly when they expressed concern about discrimination and actively applied the suggested strategies.

Another example is the study of Rokeach and Cochkane (1972), who demonstrated that highlighting contradictions between individuals’ values of equality and their racially biased attitudes led to significant reductions in prejudice – effects that persisted several weeks post-intervention (Grube et al., 1994). Similarly, studies have found that when White participants were made aware of their prejudice or when egalitarian values were activated, they tended to adjust their behavior accordingly (Monteith, 1993; Moskowitz & Li, 2011). However, individuals who are unaware of their biases are less likely to engage in change, underscoring the need to strengthen this awareness – especially in educational settings where teacher expectations can influence student outcomes. It is also important to distinguish bias awareness from other constructs, such as egalitarian motivation. While some individuals show strong efforts to be unbiased (Devine et al., 2002; Plant & Devine, 1998), even in the absence of strong bias awareness, the combination of both appears to be most effective. Notably, much of the empirical basis for this comes from US-based research, particularly on White-Black dynamics, and may need contextual adaptation when applied to other social groups or regions. Furthermore, individuals with high bias awareness tend to be more perceptive of subtle cues in their environment, indicating biased behavior, including in educational settings (Perry et al., 2015). This enhances their ability not only to reflect on their own actions but also to identify and respond to bias in others, making bias awareness a key target for intervention in professional training and teacher education.

2.3 Attribution Theory, Emotional Responses, and Individual Differences in Bias Awareness

The process of becoming aware of one’s own biases is often accompanied by a range of emotional responses, which can significantly influence individuals’ receptiveness to feedback and motivation for change. Weiner’s (1985) attribution theory offers a useful framework for understanding these responses. According to this theory, the emotional reaction to bias recognition depends on whether individuals attribute their biased behaviors to internal, controllable factors (such as personal responsibility) or to external, uncontrollable factors (such as societal norms). When individuals see their bias as controllable, they may experience guilt – a constructive emotion that can motivate efforts to change (Hofer & Waag, 2020). However, guilt is not a universal response. People with deeply ingrained biases may instead respond defensively or with anger, particularly if feedback threatens their self-concept (Perry et al., 2015).

Implicit bias is not always distressing or surprising. While some individuals feel guilt when confronted with their biases, others – especially those with lower bias awareness – may experience little or no emotional discomfort (Monteith & Voils, 1998). The acceptance of feedback about implicit biases varies according to prior attitudes and sensitivity to bias. Perry et al. (2015) emphasize that feedback is most effective when recipients are already aware of and motivated to address their biases, while it may provoke denial or resistance in those lacking such awareness.

These emotional and cognitive processes are closely linked to the motivation to change, which is central for effective bias reduction. Awareness alone is insufficient; it must be accompanied by motivation and personal commitment to translate into behavioral change.

2.4 Models of Prejudice Self-Regulation

In addition to awareness, self-regulatory strategies such as retrospective reflection can help individuals develop internal inhibition mechanisms against stereotypes (Monteith et al., 2010a; Moskowitz & Ignarri, 2009).

Theoretical models on self-regulation of prejudice stress the importance of discrepancy awareness – the recognition of a conflict between egalitarian values and biased behaviors (Monteith, 1993; Monteith et al., 2010a). Individuals with lower levels of prejudice who recognize such conflicts tend to experience guilt, which triggers reflection and compensatory behavior. Conversely, highly biased individuals often react with defensiveness or externalize blame, resulting in fewer adjustments.

Thus, both awareness and motivation to change are prerequisites for successful self-regulation. Research further shows that deeply held stereotypes are particularly resistant to change (Kumar et al., 2022), underscoring the importance of interventions that promote ongoing reflection and sustained personal commitment.

2.5 Implications for Bias Reduction Interventions

Awareness of one’s own biases is a significant predictor of increased intergroup empathy and reduced prejudice (Monteith et al., 2010a; Nelson et al., 2012; Ozier et al., 2019). This form of personal awareness is associated with the ability to internalize feedback about one’s own biases and to recognize that even subtle (rather than solely overt) biased behaviors, such as microaggressions, avoidance, or nonverbal cues, can reflect underlying prejudice (Perry et al., 2015). Previous research highlights the pivotal importance of recognizing one’s own biases (Perry et al., 2015). Based on the theoretical frameworks and empirical findings presented, it can be concluded that awareness of personal biases is a fundamental prerequisite for the effectiveness of interventions aimed at reducing them. Such awareness enables individuals to personally relate to the topics discussed, fostering sensitivity to their own biases and facilitating the adoption of self-regulatory measures to address them.

The current paper aims to build on these insights by testing the efficacy of an intervention designed to increase individual awareness of biases and provide a foundation for sustained reductions in stereotype influence. By addressing individual differences in bias awareness, this research seeks to advance understanding of how to create effective, lasting change in attitudes.

2.6 Intervention to Increase Bias Awareness

The AHA Intervention^[1] is grounded in key insights from social-cognitive theories, particularly models of attitude change (e.g., Petty & Cacioppo’s Elaboration Likelihood Model, 1986) and dual-process theories of social information processing (e.g., Fiske & Neuberg, 1990). These frameworks posit that individuals form and revise attitudes and judgments through both automatic and reflective processes. The intervention was designed to activate reflective (systematic) processing to challenge automatic, biased thinking. The AHA Intervention is delivered in a single 90-min session and combines informational input with interactive, reflective elements to promote engagement and personal relevance. It is structured around a sequence of carefully aligned phases that integrate the key components of content, relevance, and surprising moments into a cohesive learning experience. Research shows that reflection is most likely when individuals encounter personally relevant, emotionally salient, and surprising information, leading to denial, assimilation, or accommodation (Boud & Walker, 1998; Brookfield, 2017). The AHA Intervention builds on this by intentionally triggering cognitive dissonance and schema conflict through carefully chosen tasks and examples. The three core components of the intervention are: (1) creating surprising moments, (2) emphasizing the controllability and changeability of biased thinking, and (3) reinforcing the personal relevance and importance of the content.

2.7 Theoretical Foundation and Intervention Components

2.8 Study Goal and Hypotheses

The present study aims to evaluate the effectiveness of a targeted intervention designed to enhance bias awareness among pre-service teachers. Specifically, it seeks to investigate whether a brief intervention can not only increase general bias awareness but also address biases related to two critical domains: migration background and gender (binary: women/men). These domains were chosen because previous research has consistently demonstrated their significant influence on teacher judgments and actions in educational contexts. By focusing on these areas, the study aims to investigate how bias awareness can be increased in pre-service teachers. This is an important prerequisite for reducing stereotype-based misjudgments in educational practice. Building on this foundation, the study addresses several key research questions:

• Can the bias awareness of pre-service teachers be effectively increased through a short intervention?

• Are the effects of the intervention sustained over time?

• Does the impact of the intervention depend on the initial level of bias awareness in participants?

By answering these questions, the present study aims to contribute to both the theoretical understanding of bias awareness and the practical development of interventions that promote equitable teacher judgments and actions.

The combination of imparting knowledge about the unconscious influence of judgments and actions through stereotypes, with the initiation of reflection processes about one’s own (unconscious) stereotypes and prejudices, should lead to an increase in one’s own bias awareness. After the intervention, both the general bias awareness (Hypothesis 1) and the bias awareness of bias against people with a migration background (Hypothesis 2), against women (Hypothesis 3), and also against men (Hypothesis 4) should be increased. The extent to which these effects are sustainable over a period of 2–3 weeks is being explored. Furthermore, we explore whether people with a higher initial bias awareness will benefit more than people with a lower initial bias awareness, or whether people with a lower initial bias awareness benefit more. The former is conceivable, as individuals who are already open to reflecting on their biases may potentially benefit more from the intervention in line with assumptions about bias awareness. However, it is equally conceivable that they have a comparatively lower potential to further increase their bias awareness compared to individuals whose bias awareness is still less developed.

3.1 Study Design

All participants took part in the intervention with three measurement time points; there was no control group. The first measurement took place the day before the intervention and indicates the original participants’ bias awareness. The second measurement followed immediately after the intervention and should show the direct effect of the intervention on bias awareness. The third measurement was 12–35 days (M = 16.9, SD = 3.4) after the intervention and should indicate a follow-up effect of the intervention.

3.2 Participants

All participants were students of a German university and could earn participation credits, which they were required to complete as part of their study program. The sample at the first measurement includes N _T1 = 320 student teachers. The dropout rate of the second measurement was 30 participants, resulting in N _T2 = 290. The sample at the third measurement was reduced by further 99 participants. The final sample for all analyses consists of N = 191 participants (128 females, 62 males, 1 diverse) with a mean age of M = 21.46 years (SD = 3.12 years). The youngest participant was 18 years old, and the oldest was 41 years old. Twelve participants had a migration background. The mean number of semesters was M = 3.49 (SD = 3.04), ranging from 1 to 18 semesters. Nine participants provided no information about their number of semesters.

3.3 Materials

Bias awareness. To measure bias awareness, an adapted version of the scale by Perry et al. (2015) was used (Fehringer et al., 2025). This scale captures an individual’s perception of their own subtle biases. Moreover, the scale also accounts for concern about own biases. Originally, the scale was created to assess Whites’ perceptions of their own subtle biases toward Blacks in the United States. The scale consists of four items (Table 1), which can be answered on a 7-point Likert scale from “do not agree at all” to “agree completely.”

In this study, the scale was translated into German and adapted to the German context by speaking of people with a migration background instead of Black people (Table 1, t1: Cronbach’s α = 0.84, t2: Cronbach’s α = 0.81, t3: Cronbach’s α = 0.83). In addition, other variants were developed based on the original scale and used in this study. On the one hand, the general bias awareness was recorded (Table 1, t1: Cronbach’s α = 0.81, t2: Cronbach’s α = 0.79, t3: Cronbach’s α = 0.79) and bias awareness of gender prejudices (Table 1, against women, 4 items, t1: Cronbach’s α = 0.81, t2: Cronbach’s α = 0.77, t3: Cronbach’s α = 0.80, and against men, 4 items, t1: Cronbach’s α = 0.80, t2: Cronbach’s α = 0.78, t3: Cronbach’s α = 0.81). To capture the complexity of bias awareness across different social groups, we assessed bias awareness separately for women and men, as gender-related biases can vary depending on context and target group.^[2] For the migration background, a broader category was used to avoid excessive fragmentation.^[3] Additionally, a general bias awareness scale was included to measure participants’ overall sensitivity to biases beyond specific social categories. This approach allowed us to examine both specific and general aspects of bias awareness relevant to the school context.

3.4 Procedure

As part of the present study, prospective teachers were asked about their own biases. On the one hand, the awareness of the influence of stereotypes and prejudices, in general, was measured, and, more precisely, the influence in relation to stereotypes and prejudices against people with a migration background or according to gender. The order of these four scales was the same for each participant: (1) General, (2) Migration Background, (3) Female, and (4) Male. In addition, a short intervention was conducted aimed at raising awareness of one’s own biases. To evaluate the effectiveness of the intervention, a survey design with three measurement times (a pre-measurement, a post-measurement directly after the intervention, and a post-measurement about 14 days after the intervention) was chosen. Student teachers were invited via a university mailing list and social media to participate in this study. After prior registration for the desired dates for a 90-min input session (the short intervention), the link to the first survey was sent to the participants. The input part took place 1–3 days after the first survey (t1) digitally via Zoom (cloud-based video conference service). Immediately after the input part, the students conducted the first follow-up survey (t2). After 10 days, the participants were asked by email to participate in another follow-up survey (t3). Both the pre-measurement (pre-survey before the intervention) and the post-measurements took about 15 min and were carried out using an online questionnaire in the UNIPARK survey software.

The timeline of the intervention study was as follows:

1. Preliminary survey of 15 min (1–3 days before the intervention)

2. 1 × 90 min input session via Zoom (intervention)

3. Follow-up survey of 15 min (immediately after the intervention)

4. Follow-up survey 2 of 15 min (approx. 14 days later)

Informed consent has been obtained from all individuals included in this study.

3.5 Analyses

Before the hypotheses were tested, a dropout analysis was conducted by four repeated measurement analyses of variance (ANOVAs), one for each bias awareness variable (i.e., general, migration background, female, male). These ANOVAs compared the two dropout groups, after the first and second measurements, as well as the group that finished the third measurement. The considered variable was always from the first measurement.

The effectiveness of the intervention (Hypotheses 1–4) was examined using repeated-measures analyses of covariance (ANCOVAs), conducted separately for each of the four BA scales comparing the three measurements and controlling for gender and migration background. The follow-up analyses contrast the first measurement (before the intervention) against both measurements after the intervention as well as consider all three pairwise post hoc comparisons, i.e., between the first and the second as well as the third measurement and between the second and the third measurement. The sustainability of the effects over a period of 2–3 weeks was exploratively analyzed by the post hoc comparisons between the first and the third measurement. The second explorative analysis considers the influence of the initial bias awareness on the effects and was analyzed for each BA scale by linear regression analyses. Additionally, the 95% confidence intervals for the coefficients were estimated using bootstrapping. All statistics were performed using R statistics, version 4.1.2 (R Core Team, 2021), using the package afex, version 1.3.0 (Singmann et al., 2023), emmeans, version 1.8.7 (Lenth et al., 2023), psych, version 2.1.9 (Revelle, 2021), and boot, version 1.3-28 (Canty et al., 2021).

4.1 Bias Awareness (General)

The ANCOVA showed a significant difference between all three measurements, F(2, 380) = 19.01, p < 0.001, η ² = 0.02, and a significant contrast between the first measurement and both measurements after the intervention, t(380) = 4.05, p < 0.001 (Hypothesis 1). However, only the post hoc test between the first and the second measurement revealed a significantly increased bias awareness, t(380) = 5.83, p < 0.001. The follow-up effect between the first and the third measurement was not significant, t(380) = 1.18, p = 0.465. The post hoc comparison between the second and third measurement confirmed a significant decrease in BAS values, t(380) = 4.65, p < 0.001 (see also Figure 1).

The linear regression model demonstrates a negative relationship between the initial BAS value before the intervention and the increasing effect of the intervention measured as the increase from the first measurement to the second measurement, b = −0.36 [−0.47; −0.26], t(189) = 7.33, p < 0.001, adj. R ² = 0.22.

4.2 Bias Awareness (Migration Background)

According to Hypothesis 2, the BA scale for migration background differed between the three measurements, F(2, 380) = 32.64, p < 0.001, η ² = 0.03, with a significant increase from the first measurement to the second and third measurement combined, t(380) = 6.78, p < 0.001. There was a significant increase from the first to the second measurement, t(380) = 8.07, p < 0.001, and contrary to the general scale of the BA, there was also a significant increase from the first to the third measurement, t(380) = 3.67, p < 0.001. However, similar to BA (general), there was a significant decrease from the second to the third measurement, t(380) = 4.40, p < 0.001 (see also Figure 1).

There was a significant negative relationship between the initial BAS values (at the first measurement) and the effect of the intervention (from the first to the second measurement), b = −0.28 [−0.38; −0.18], t(189) = 5.68, p < 0.001, adj. R ² = 0.14.

4.3 Bias Awareness (Female)

Hypothesis 3 could also be confirmed for the BA (female) scale. The ANCOVA showed significant differences between the three measurements, F(2, 380) = 14.35, p < 0.001, η ² = 0.02, and a significant contrast between the first measurement and both measurements after the intervention, t(380) = 5.26, p < 0.001. Also, both post hoc tests comparing the first measurement with the second and third measurements, respectively, were significant, t(380) = 5.07, p < 0.001, and t(380) = 4.04, p < 0.001, indicating an increase from the first measurement. In contrast to BA (general/migration), the level stayed similar from the second to third measurement, t(380) = 1.03, p = 0.560 (see also Figure 1).

This BA-scale showed the descriptively largest negative relationship between the first measurement and the effect (from the first to the second measurement, Hypothesis 2), b = −0.49 [−0.63; −0.36], t(189) = 7.24, p < 0.001, adj. R ² = 0.21.

4.4 Bias Awareness (Male)

Similarly to the other scales, Hypothesis 4 could be confirmed with a significant ANCOVA, F(2, 380) = 17.14, p < 0.001, η ² = 0.02, and a significant contrast between the first measurement and the second and third measurement, t(380) = 5.15, p < 0.001. Also, both post hoc comparisons between the first and each of the other measurements were significant, t(380) = 5.85, p < 0.001 (with the second measurement), and t(380) = 3.06, p = 0.007 (with the third measurement), indicating an increase from the first measurement. In contrast to BA (female), there was a significant decrease from the second to the third measurement, t(380) = 2.79, p = 0.015 (see also Figure 1).

Also, there was a significant negative relationship between the initial BAS values at the first measurement and the effect (increase from the first to the second measurement, Hypothesis 2), b = −0.33 [−0.44; −0.23], t(189) = 5.95, p < 0.001, adj. R ² = 0.15.

5.1 Limitations

The results of our study support the assumption that prospective teachers with lower bias awareness benefit the most from the given intervention. However, it is important to emphasize that the effectiveness of interventions to increase bias awareness depends heavily on the willingness and openness of participants to accept and apply new concepts (Kumar et al., 2022). Individual differences and context-specific factors may play an important role in assessing how effective and sustainable these interventions can be for people with low bias awareness.

The present study lacks a control group and only considers a pre-post comparison, weakening the interpretation of the results (Aggarwal & Ranganathan, 2019) since the cause(s) for the changes could be others than the intervention. However, it seems implausible that all (or at least most) participants were exposed to the same (similar) “events,” reducing bias by chance during the study. Furthermore, the intervention targets well-documented theoretical constructs (Monteith et al., 2010a; Perry et al., 2015), and the outcomes align with existing literature, reinforcing the plausibility of the findings.

A selective group of people could also have taken part in the intervention, which would limit the generalizability of the results (Cook et al., 2002; Heckman, 1979). However, it should be noted that the study was announced very openly, and the participants were therefore not aware of the topic in advance. They were only informed that they would be taking part in a three-part study with an input session on a topic relevant to prospective teachers. The fact that the student teachers have additional incentives to participate regardless of the content due to the allocation of test subject hours, which all prospective teachers have to complete as part of their studies, also limits the problem that only very motivated, committed students may have taken part.

The sample size shrank from t2 to t3. This drop-out could be problematic if the drop-out is selective (Stellato et al., 2025) and, for example, all motivated or possibly already sensitized participants continue to participate, while other people drop out. A different performance of the groups in the post-measurement could then be partly due to the higher motivation or sensitization. The motivation of the participants was not recorded in this study. However, age and different bias awarenesses (separate for each group) at the three measurements as a selection factor could be ruled out. These results suggest that the dropout rate is completely at random, which would not affect the performed analyses (Stellato et al., 2025).

While our study assessed bias awareness toward migration background in a general manner, future research should consider the heterogeneity within migrant groups, as attitudes and biases may differ significantly depending on specific cultural or ethnic backgrounds.