2026-04-14
Why Bias Training Does Not Work the Way We Want It to, and What to Do Instead
Kaisa Vaittinen
Decades of research show one-off bias training rarely changes behaviour — and sometimes makes things worse. Here is what the evidence supports instead.
One of the biggest investments in the human resources field this millennium has been unconscious bias training, the attempt to reduce unconscious prejudices through psychoeducation. Organisations in the United States spend around eight billion dollars a year on this. The logic is simple: if people understand the mechanisms of their own prejudice, they will be able to make fairer decisions. Understanding changes behaviour. That is what we usually believe.
The problem is that research does not really support this assumption. Bias training produces knowledge and some attitudinal change in the immediate context of the training itself, but these changes do not usually hold over time and do not transfer into behaviour or into organisation-level change. At worst, the training produces a backlash.
What this essay argues
- Single-session unconscious bias training shows minimal long-term behavioural change.
- Mandatory training can increase resentment and entrench bias.
- Structural changes (hiring, promotion, accountability) outperform attitude work.
- Sustained, contextual practice and feedback beat awareness alone.
- The popularity of bias training reflects organisational comfort, not effectiveness.
What does the research say?
Frank Dobbin and Alexandra Kalev have studied the effects of diversity programmes for over three decades, covering more than 800 organisations. Their central finding is that traditional bias training does not increase the representation of women or minorities at the leadership level, and in some cases it even decreases it. Training experienced as mandatory activates psychological reactance.
Kalinoski et al. (2013) found moderate effects on attitudes when training is well designed, but effects on behaviour remain small or non-existent. Forscher et al. (2019) confirmed that implicit prejudice measurements change short-term but do not predict behaviour changes. Paluck and Green (2009) concluded that the causal effects of many widely used interventions are unknown due to weak methodology.
Why do bias trainings not necessarily work?
First, understanding and behaviour are not the same thing. Second, mandatory training activates psychological reactance. Third, bias is not an individual problem but a structural one — it lives in the shared norms of the work community.
What to do instead?
Dobbin and Kalev have identified practices that demonstrably shift representation: formal mentoring programmes, explicit naming of diversity responsibility at leadership level, and structural changes to recruitment processes. Iris Bohnet's research on behavioural design confirms the same: change the environment, do not try to change the people.
The affective dimension
There is, however, one dimension that structural solutions do not reach and that cognitive training does not touch. Research has consistently found that implicit attitudes which have been formed on an emotional basis change more effectively through affective interventions than through cognitive ones (Edwards, 1990; Fabrigar & Petty, 1999). In practice this means that knowledge of the existence of one's prejudices is not enough. This is because prejudices are not primarily cognitive but emotional structures. They are felt in the body before they become thoughts. For bias training, this is good news: training can be supplemented with components that produce genuine emotional reactions in participants and give them the opportunity to observe those reactions in real time. In other words, the participant themselves reacts to a situation genuinely and is given the chance to examine their own reaction before it has time to turn into an interpretation or a defence. Creating such a situation may not be possible from within the group, but may require an outsider who can function as a stimulus and to whom one can react safely.
What does this mean in practice?
Removing bias training is not the same as giving up on diversity work. Resources can be redirected to structural practices that work. Structural intervention requires leadership commitment to process changes, measurements, and the acknowledgement that the problem is in the organisation's structures, not in individual employees' prejudices.
The schema beneath the surface
Organisations carry collective shared schemas about who "looks like" the right kind of leader, employee, or colleague. These schemas live in everyday conversations and countless small decisions. This layer is not reached by training, nor always by structural change. It only comes into view when someone makes it visible.
Bias training asks: "are employees aware of their prejudices?" Structural change asks: "are our processes such that they neutralise the effects of prejudice?" The third question: "what does our organisation actually treat as normal, reward, and punish, below the level that can be measured?"
The purpose of this text is not to declare bias training useless. If we are investing eight billion dollars a year in something that does not produce effectiveness, it is honest to ask why, and what could be done instead.
If this text produces frustration with current solutions, that is ok. If it produces a defensive reaction, that is also ok. Both are exactly the work that has to be done before one can actually start to change what needs to be changed.
Sources and background reading
On the effectiveness of bias training
- Dobbin, F. & Kalev, A. (2016). Why diversity programs fail. Harvard Business Review, 94(7–8), 52–60.
- Dobbin, F. & Kalev, A. (2018). Why doesn't diversity training work? Anthropology Now, 10(2), 48–55.
- Kalinoski, Z. T. et al. (2013). A meta-analytic evaluation of diversity training outcomes. Journal of Organizational Behavior, 34(8), 1076–1104.
- Forscher, P. S. et al. (2019). A meta-analysis of procedures to change implicit measures. Journal of Personality and Social Psychology, 117(3), 522–559.
- Paluck, E. L. & Green, D. P. (2009). Prejudice reduction: What works? Annual Review of Psychology, 60, 339–367.
- Edwards, K. (1990). The interplay of affect and cognition in attitude formation and change. Journal of Personality and Social Psychology, 59(2), 202–216.
- Fabrigar, L. R. & Petty, R. E. (1999). The role of the affective and cognitive bases of attitudes. Personality and Social Psychology Bulletin, 25(3), 363–381.
- Paluck, E. L., Porat, R., Clark, C. S. & Green, D. P. (2021). Prejudice reduction: Progress and challenges. Annual Review of Psychology, 72, 533–560.
On behavioural design and structural solutions
- Bohnet, I. (2016). What Works: Gender Equality by Design.
- Dobbin, F. & Kalev, A. (2022). Getting to Diversity: What Works and What Doesn't.
On psychological reactance
- Brehm, J. W. (1966). A Theory of Psychological Reactance.