Which of the following do you think has the greatest impact when you want to change people’s behavior? The behavior could be anything from brushing your teeth to getting vaccinated. How would you rank them, from most to least effective?

  1. knowledge
  2. behavioral skills, such as practicing how to refuse a cigarette
  3. general skills, such as self-control
  4. attitudes toward the behavior itself
  5. emotions, such as fear of a health risk
  6. habits
  7. beliefs
  8. monitors and reminders
  9. social support
  10. access, such as having a vaccination site close to home

Hold on to your ranking, and let’s see what a study found by reviewing multiple meta-analyses from various domains such as health, finance, and environment!

Albarracín and colleagues (2024) gathered meta-analyses on 16 factors, including the ten in the quiz. For each factor, the authors compared, where available, how strongly the factor correlates with behavior and how much interventions aimed at the factor actually changed behavior. They then sorted the factors into four groups by the average size of the intervention effects: negligible, small, medium, and large.1

Negligible effects

1. Interventions that target knowledge deficits have a negligible effect on behavior.

These interventions include education and other didactic approaches intended to reduce a knowledge deficit. In a meta-analysis of vaccination interventions, for example, neither providing general information (OR = 1.04) nor correcting misinformation (OR = 0.94) increased vaccination uptake.

Financial literacy, for another example, has a medium correlation with desirable financial behaviors (r = 0.29), but only a negligible one when literacy is measured first and behavior afterward (r = 0.09). In other words, knowing more about money does little to predict how well people manage money later. The larger correlation may simply reflect that people who already manage money well tend to know more about money.

2. Training general skills, such as controlling attention or resisting temptation, has a negligible effect on behavior overall.

Many behavior change programs train these broad skills on the rationale that they help people control undesirable behaviors. However, executive functioning skills, such as inhibitory control, have only a small correlation with specific behaviors such as disinhibited eating (r = −0.14).

3. Changing general attitudes and beliefs, including mindsets, has a negligible effect as well.

Growth mindset is a familiar case for many educators, and interventions have been developed to change people’s mindsets. However, a meta-analysis of these interventions found a negligible effect on behavior, both in experiments that successfully changed mindset (d = 0.04) and across all experiments (d = 0.05).

Consequently, learning designers who routinely create curricula to raise awareness and understanding, or to improve general skills and attitudes, may be disappointed if they expect enlightened learners to choose the correct course of action by these interventions alone. Learners may come away informed but behaviorally stagnant.

Small effects

Even the factors that do change behavior mostly have small effects, so a program may need to address several of them at once.

1. Train specific behavioral skills through modeling, practice, and feedback.

Behavioral skills interventions teach a specific behavior, such as refusing an invitation to smoke or drink. The overall effect of these interventions is small, but the authors consider behavioral skills one of the more promising factors because their effects are larger than those of general skills. Most of the components of behavioral skills interventions are familiar to our learning designers: demonstrating the skill, practice, feedback, review, and additional practice.

2. Show people what others actually do, and use role models who resemble them.

Interventions that target descriptive norms, meaning what others actually do, have small effects overall. Role models had a medium effect (d = 0.51), and interventions delivered by facilitators who demographically resemble the people they teach were more successful at increasing condom use than those delivered by facilitators who did not. In a training program, this finding may mean choosing facilitators whom learners see as similar to themselves, featuring peers from the learners’ own roles in examples and videos, and, when the desired behavior is already common, letting learners know that most of their colleagues already do it.

3. Add reminders and tracking tools, preferably both manual and digital ones.

Manual reminders, such as tracking sheets and paper planners, had a small effect on colorectal cancer screening (OR = 1.85), and combining them with computer-generated reminders had a medium effect (OR = 2.57). The authors suggest reminders may be most useful in combination with interventions that address other factors.

Medium effects

1. When the goal is to stop a habit, train people to notice its cues and respond differently.

Among the individual factors, habit interventions show the strongest effects on behavior. All of the figures the authors report, however, come from curbing existing habits. Habit reversal training for tics, for example, teaches patients to notice what triggers a tic and to respond with an incompatible movement.

The authors also point out a limitation of habit interventions: people have to perform a behavior before it can become automatic. This suggests that, for a new behavior, building the habit may be most useful as a later step, once people have the skills, access, and support they need to perform the behavior in the first place.

2. Give people social support for the specific behavior.

Social support interventions, which often take the form of support groups, have medium effects overall. Unlike norms, social support is tied to a particular goal: it means others are willing to give advice or other help with that goal.

Training research points in the same direction. In a meta-analysis of training transfer, trainees with a more supportive work environment tended to apply more of what they learned on the job, especially for open skills such as leadership (Blume et al., 2010). In a training program, building social support may mean involving learners’ managers and peers and forming peer cohorts.

Large effects

Remove barriers and make the desired behavior easy to do.

Access has the largest effects in the review, even though its correlation with behavior is small. Providing transportation to vaccination sites, or placing sites close to potential users, doubled vaccination coverage.

Arthur and colleagues (2003) make a related point about training: trained skills will not show up as job behavior if employees do not have the opportunity to use them, although their meta-analysis did not test this directly. Before designing a training program, it may be worth asking whether learners will have the time, tools, and resources to act on what they learn.

What does this mean for designing educational or training programs?

Information still has a place. The authors recommend providing basic information when an audience is unfamiliar with a behavior or its purpose. However, when a program aims to change behavior, information is only one part of the design. Whether people perform the behavior may also depend on opportunities to practice it, support from others, timely reminders, and access to the resources they need.2

References

Albarracín, D., Fayaz-Farkhad, B., & Granados Samayoa, J. A. (2024). Determinants of behaviour and their efficacy as targets of behavioural change interventions. Nature Reviews Psychology, 3(6), 377–392. https://doi.org/10.1038/s44159-024-00305-0

Arthur, W., Jr., Bennett, W., Jr., Edens, P. S., & Bell, S. T. (2003). Effectiveness of training in organizations: A meta-analysis of design and evaluation features. Journal of Applied Psychology, 88(2), 234–245. https://doi.org/10.1037/0021-9010.88.2.234

Blume, B. D., Ford, J. K., Baldwin, T. T., & Huang, J. L. (2010). Transfer of training: A meta-analytic review. Journal of Management, 36(4), 1065–1105. https://doi.org/10.1177/0149206309352880

  1. Cohen’s d and Hedges’ g compare the average behavior of people who received an intervention with that of people who did not, in standard-deviation units, where 0 means no difference. A correlation (r) shows how closely a factor and a behavior go together, from −1 to 1, where 0 means no relationship. An odds ratio (OR) compares the odds of the behavior in the two groups, where 1 means no difference. By the authors’ benchmarks, an effect counts as negligible below d = 0.2, r = 0.1, or OR = 1.44; small up to d = 0.49, r = 0.23, or OR = 2.47; medium up to d = 0.79, r = 0.36, or OR = 4.26; and large above that. ↩

  2. The outcomes in Albarracín and colleagues’ review are behaviors such as vaccination, recycling, and work behavior. The authors also caution that some behaviors, populations, and contexts might be unique, so no review can predict the result of an intervention in every setting. ↩