1. Home
  2. Chapter 11
Trippas, D. & Pachur, T. (2019). Ways to learn from experience. In R. Hertwig, T. Pleskac, T. Pachur, & the Center for Adaptive Rationality (Eds.), Taming uncertainty (pp. xx–xx). Boston, MA: MIT Press. doi:XXXXXXX

Chapter 11 Ways to Learn From Experience

In this interactive element for Chapter 11, you can experience the predictions of the rule-based and the exemplar-based strategies in different environments—and thus explore the ecological rationality of these strategies. In particular, the goal is to demonstrate how the rule-based and the exemplar-based strategies differ in extrapolation, that is, the ability to correctly judge the criterion value of new objects that have criterion values that are more extreme than those experienced in a previous training phase. You can contrast two environments: a linear environment, in which the criterion values of the different objects are a linear additive function of the cue values; and a nonlinear environment, in which the criterion value of the objects follows from the cue values in a nonlinear, cubic fashion. The cue profiles of the different objects and their criterion values in the two environments are shown in the table below. Importantly, you can also vary which and how many objects of the environment are assumed to be included in the training phase—and which serve to inform the parameters of the strategies about the structure of the environment.

Criterion value
# ObjectCue 1Cue 2Cue 3Cue 4Use in trainingLinear environmentNonlinear environment
1111110.16
21110.9.47
31101.8.71
41100.7.88
51011.7.88
61010.6.97
7100151
81000.4.94
90111.6.97
10011051
110101.4.94
120100.3.82
130011.3.82
140010.2.62
150001.1.35
16000000

Linear environment

Rule-based strategy

Exemplar-based strategy

Parameters of the rule-based strategy
K: 0 W1: 0.4 W2: 0.3 W3: 0.2 W4: 0.1
Parameters of the exemplar-based strategy
H: 18.65 W1: 0.22 W2: 0.27 W3: 0.27 W4: 0.24

Nonlinear environment

Rule-based strategy

Exemplar-based strategy

Parameters of the rule-based strategy
K: 0.74 W1: -0.09 W2: 0.16 W3: -0.02 W4: -0.05
Parameters of the exemplar-based strategy
H: 19.69 W1: 0.28 W2: 0.23 W3: 0.26 W4: 0.23
×

In contrast to the rule-based strategy, the exemplar-based strategy is constrained in its ability to accurately predict objects with criterion values that lie outside the range of criterion values that were encountered during the training phase. Conversely, the rule-based strategy is highly constrained to capture the cur-criterion relationship in a nonlinear environment, whereas the exemplar-based strategy can represent any statistical structure encountered during training. This illustrates the different ecological rationalities of the rule-based and the exemplar-based strategies.

You have earned a badge! Click me to read more.

Click on any reward badge in the footer to review take home messages of this or other elements.