Eldad Yechiam

Eldad Yechiam, Ph.D
Technion - Israel Institute of Technology

I am an Associate Professor of Behavioral Sciences at the Technion - Israel Institute of Technology. My research focuses on individual differences in decision making and decision making processes at the individual level. This research area has applications in Clinical Science, Artificial Intelligence, and Human Factors Engineering. See examples below.

Research Topics:

1) Individual differences in decision making

My main experimental activity in the last six years has been in modeling individual differences using reinforcement learning algorithms; and devising the methodology of applying these models to the individual case. Traditionally, these models have been applied to predict group averages (see e.g., Roth & Erev, 1995). I have studied the use of these models at the individual level, and have developed new statistical techniques for model assessment. These techniques include a method for assessing the generalizability of model predictions that is also useful for assessing the adequacy of tasks for evaluating individual differences (see Yechiam & Busemeyer, 2008).

Another series of papers demonstrate the usefulness of models in assessing the diversified reasons leading to impairments in decision making. In a study in Psychological Science (Yechiam et al., 2005), we have shown that what appears as similar decision making impairments on a task known as the Iowa Gambling task (Bechara et al., 1994) can actually be due to substantially different component processes. This resulted in a “map” of the difference in cognitive style between different neuropsychological populations.

A part of the outline of neuropsychological populations. The X axis denotes the weight of gains compared to losses (right – more gain oriented). The Y axis denotes the extent of recency. The bubble size denotes the choice consistency. From Yechiam et al. (2005).

In the most recent study on this issue (with my past student Eyal Ert), we use behavioral methods to complement modeling and examine the decision making traits that are consistent in individuals’ behavior. Our results pose important limitations in the constructs highlighted by Kahneman and Tversky (1979) in their seminal paper. First we show that across the gain and loss domains people exhibit positive consistency, inconsistently with the interpretation of risk as diminishing sensitivity to payoff differences from zero. Second, the results show consistency in responding to gains and losses only when a risky alternative with gains and losses is compared to a constant outcome. These findings denote an important deviation between population models of decision making and consistent latent constructs within the individual, and specifically, suggest the importance of a construct labeled as risk acceptance (and defined as sensitivity to strong signals of risk such as in deciding between constant versus probabilistic outcomes).

Correlation between risk taking in the gain and loss domain. The term AU is for Avoidable uncertainty (e.g., get 10 versus a gamble between 0 or 20. The risk is avoidable by selecting the safe alternative of getting 10). From Yechiam & Ert (2011).

2) Experience based decisions

Ido Erev and his colleagues had discovered that people overweight small probability events in description based tasks, but underweight these events in experience based tasks (people behave as if they believe that rare events will happen to them less than they do; Barron & Erev, 2003). These tasks are typically used as models of naturally occurring behaviors. I have studied the implications of a related but under-investigated class of decision tasks, involving foregone payoffs (i.e., when all payoffs of the available alternatives are revealed upon one’s choice of any alternative in a repeated choice task). The findings reveal that in this condition decision makers show super-underweighting of small probability events, even compared to other “experience based” decision tasks (e.g., Yechiam & Busemeyer, 2005; 2006). A related finding was that a major dimension of individual differences in decision making that is impaired in drug abusers, involves risky behavior when there are distracters in the form of foregone payoffs (Yechiam et al., 2005). This implies that for understanding the decision making style implicated in addictive behavior one needs to assess not only basic decision style, but rather a situation that taxes working memory and self control to some extent.

3) Effect of losses

My most recent interest in the last two years involves the effect of losses. There is an ongoing debate in the area of decision making regarding the existence and extent of “loss aversion”, the tendency of negative events to have more subjective impact than positive ones. My contribution involves an attempt to clarify the apparent paradox that losses lead to increases in arousal (e.g., increases in heart rate and pupil dilation) even when behaviorally individuals do not respond to losses differently from gains. While demonstrating this inconsistency at the individual level, I have highlighted a reasonable explanation for it. Under the attention based model of losses (Hochman & Yechiam, in press; Yechiam, submitted) losses signal an important situation for the organism’s immediate survival, and an alternative producing losses becomes a more focal center of attention. However, all outcomes associated with this alternative become subjectively more significant, even gains that are associated with it. This leads to no asymmetry in the weighting of gains and losses in mixed gambles.