Decisions from Experience vs. Learning: Which Knows Best?

Decisions, decisions, decisions. Sometimes we have been lucky and we have been aided in this process: we have learned about these decisions before in theory, or we have been through something similar (enough) before and have got some experience under our belt to help us out. Bruce Rigal explains which one might help you out more!

Hertwig et al. (2004) look at two different methods by which people make decisions in risky situations. Most difficult decisions we make are risky in the sense that we do not know, ex-ante, which decision will maximise our happiness (or ‘utility’ in Behavioural Economics terminology).

The experimenters look at simple decisions involving two options. The subject will naturally want to know the possible outcomes (payoffs) of each of these choices before making a decision. The first method of deciding is called ‘decision by description.’ So the experimenter might ask which is preferred of these two options:

A: Get £4 with probability 0.8; £0 otherwise. Or

B. Get £3 for sure.

This is the traditional way we ‘teach’ students in school or University. It is important to note that only humans can learn this way (as far as we know!).

Kahnemann and Tversky (1979 and 1992) developed Prospect Theory (Original and Cumulative, respectively) to describe and predict how humans make such risky decisions by description. They found that in such cases people overweight small probabilities and underweight large probabilities. In the above example this would push them to choose options like B (that provide certainty) more than what might be considered ‘rational’ because they overweight the low probability of receiving zero in choice A.

The second way we humans learn, in common with other animals, is through ‘decision by experience’. In this case we are not told the underlying distribution but rather explore (sample) what actually happens when we choose A and B and then by seeing the outcomes of our choices we deduce the underlying distribution of payoffs. Think about how a bird decides which orchard has the best fruit: the bird will ‘explore’ the options and experience the outcomes until it intuitively knows which orchard is the best to visit.

Hertwig et al. (2004) compare the results when subjects decide based on the two methods. They discover that in decisions from experience subjects underweight small probabilities and overweight large ones while in decisions from description low probability outcomes have more impact than they ‘deserve’ (as Kahnemann and Tversky found). They posit a number of reasons for this result, including limited information search and recency effects.

What does this have to do with teaching children and adults? It seems self-evident that one of the main purposes of education is to teach students how to make decisions that involve risk. For instance, in the business school we often teach students how to make an investment decision among different risky alternatives. As teachers we are being told that students are more engaged and thus learn more when they are learning from experience (interactive classrooms) than from description (traditional teaching). If an important goal of education is to help our students make better risky decisions then we should consider teaching using both methods so the two biases (underweighting and overweighting small risks) can help offset each other. This will encourage our students to make better decisions in the classroom and in the ‘real world’.


Hertwig, R., Barron, G., Weber, E. U., & Erev, I. (2004). Decisions from Experience and the Effect of Rare Events in Risky Choice. Psychological Science, 15(8), 534–539.