Research
Go to: Publications , Data Sets.
I think of research as a fun hike in “assumption hills” overlooking the “land of behaviors.”
The main goal of this hike is to find good observation points that facilitate the discovery of interesting regularities in the land of behaviors (Erev, 2020).
​
​
​
​
​
​
​
​
​
​
My favorite hill is defined by three main assumptions:​
A1. Money makes the world go round.
A2. Good predictions of the direction of the spin (the impact on money) can help design mechanisms that reduce social conflict and improve social welfare.
A3. In certain settings, traditional economic analysis that assume rational choice provides useful predictions of the impact of money.
The view from my favorite hill reveals several interesting observations.
Six examples are listed below:
​
1. Contingent Rationality
When the payoff maximizing strategy also maximizes the probability of success, simple rational models provide useful predictions of human behavior
4. Laziness and the Wavy Recency Effect
There are two very different contributors to the descriptive value of the reliance on small sample assumption reveals. When people can sample new information, the main driver is laziness
​
(Hertwig et al., 2004; Erev & Plonsky, 2023).
In contrast, when people depend on on their memory, the results suggest that the underlying processes are rather sophisticated and reflect an effort to rely on the most similar past experiences. This effort approximates the optimal Baysian strategy in dynamic settings, but implies reliance on small samples, and a wavy recency effect, in static settings
(Plonsky et al., 2015; Plonsky & Erev, 2017).
The sophistication suggests that the underlying processes are more similar to machine learning classification algorithms than to simple heuristics
2. The Description-Experience Gap
When the condition considered in Contingent Rationality does not hold (i.e., when the maximizing strategy does not maximize the probability of success), the direction of the deviations from rational choice is highly sensitives to the availability of feedback.
Our results reveal a robust description-experience gap: Before receiving feedback (when deciding based on descriptions of the incentive structure) people tend to overweight rare and extreme outcomes, and feedback reverses this bias and triggers underweighting of rare events
(Barron & Erev, 2003; Hertwig et al., 2004; Erev et al., 2017).
5. Gentle Rule Enforcement, and Big Data without Big Brothers
One of the optimistic implications of the observations listed above is the suggestion that it is often possible to enforce rule without the use of severe punishment
(Erev et al., 2010c; 2010d),
and with limited invesion of privacy
​
3. Reliance on Small Samples
The results summarized by Contingent Rationality and The Description-Experience Gap can be captured with simple cognitive models assuming reliance on small sample of similar past experiences. The value of models that share this assumption was clarified in a series of choice prediction competitions
6. The J/DM Separation Paradox
The main difference between my favorite hill, and the hills preferred by early behavioral decision scientists, involves the assumed impact of separating judgment from decision making. Early research studies judgment and decision making separately, and finds oversensitivity to rare events in both. The view from my hill suggests that without the separation people exhibit the opposite bias: Insufficient sensitivity to rare events. This J/DM Separation Paradox can be explained with the assertion that the separation requires explicit presentation of the rare events, and this presentation increases the attention they receive
​