Adolescents display distinctive tolerance to ambiguity and to uncertainty during risky decision making

Abstract

Although actuarial data indicate that risk-taking behavior peaks in adolescence, laboratory evidence for this developmental spike remains scarce. One possible explanation for this incongruity is that in the real world adolescents often have only vague information about the potential consequences of their behavior and the likelihoods of those consequences, whereas in the lab these are often clearly stated. How do adolescents behave under such more realistic conditions of ambiguity and uncertainty? We asked 105 participants aged from 8 to 22 years to make three types of choices: (1) choices between options whose possible outcomes and probabilities were fully described (choices under risk); (2) choices between options whose possible outcomes were described but whose probability information was incomplete (choices under ambiguity), and (3) choices between unknown options whose possible outcomes and probabilities could be explored (choices under uncertainty). Relative to children and adults, two adolescent-specific markers emerged. First, adolescents were more accepting of ambiguity; second, they were also more accepting of uncertainty (as indicated by shorter pre-decisional search). Furthermore, this tolerance of the unknown was associated with motivational, but not cognitive, factors. These findings offer novel insights into the psychology of adolescent risk taking.

Figure 1. Experimental tasks.

(A) Choices under risk were represented by a wheel of fortune consisting of 10 slices of different colors. The orange slices were associated with possible earnings or losses; if the spinner stopped at the blue slices, the player would receive nothing. (B) Choices under ambiguity were represented in exactly the same way as the risky gambles, but this time (gray) occluders on top of the wheels hid part of the information. (C) Three levels of ambiguity were implemented. (D) In the choices under uncertainty task, participants were able to sample from two payoff distributions before making a final choice. Participants sampled by pressing the sample button and chose by clicking on the corresponding circle. All experimental tasks were self-paced.

Figure 2. Behavioral and modeling results for risk and ambiguity attitude.

Parameter estimates of risk (α) and ambiguity (β) attitudes. The parameter values were transformed such that, in each graph, values larger than zero indicate risk/ambiguity seeking and values smaller than zero indicate risk/ambiguity aversion.

Figure 3. Sample sizes in choices under uncertainty.

Mean number of samples per age group and separately for the gain and loss domain (error bars show standard errors).

Figure 4. Stopping rules.

Percentage of times participants stopped after encountering a positive outcome (i.e., a nonzero outcome in the gain domain or a zero outcome in the loss domain), binned by age group. White diamonds represent group means.