Thinking Through Complexity

How to cite

Rasocha, Vlastimil; Bernheim, B. Douglas; Hartley, Jon. 2026. "Thinking Through Complexity." AEA RCT Registry. AEARCTR-0017632. https://doi.org/10.1257/rct.17632-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Vlastimil Rasocha* , B. Douglas Bernheim , Jon Hartley

* Primary investigator

Registry ID
AEARCTR-0017632
DOI
https://doi.org/10.1257/rct.17632-1.0
First registered
Last updated
Status
In development
Trial dates
2026-02-15 – 2026-12-31
Countries
Private
JEL codes
D9, C91
Registry record
http://www.socialscienceregistry.org/trials/17632

Abstract

This study investigates how individuals reason in complex economic environments involving indirect and feedback effects. Participants in an online experiment take part in a “transfer game” with three fictitious players and a fixed set of rules that determine how money moves between players over two or three rounds. In each game, participants predict how many coins one player will end up with and earn a bonus based on prediction accuracy. A key feature of the design is that participants do not see the full set of transfer rules directly. Instead, they can consult a simple calculator that reveals the consequence of a specific transfer rule they choose to look up. By observing both participants’ predictions and the information they choose to look up, we study how people simplify complex environments by focusing on some effects and ignoring others, and how these simplifications depend on the complexity of the environment, information costs, and incentives for accuracy.

Experimental design

This study is an online lab experiment in which participants complete a “transfer game” prediction task. Each game involves three fictitious players (Blue, Red, Yellow) with bank accounts. Yellow starts with 100 coins and the other players start with 0. The game lasts either two or three rounds, depending on treatment. In each round, each player follows a rule that specifies how many coins they transfer to each of the other players, how many they keep, and how many they spend; the rules change by round.

Participants’ task is to predict how many coins will be in Blue’s account at the end of the game. Participants do not see the underlying rules directly. Instead, they can use a custom calculator that reveals the transfers implied by a specific rule they choose to look up (by selecting the round, sender, receiver, and a starting balance). We record participants’ final predictions and the complete history of their calculator look-ups.

The experiment uses a between-subject design with randomized variation in: (i) the number of rounds in the game (two vs. three), (ii) whether calculator look-ups are free or carry a small per-look-up cost, and (iii) the size of the accuracy-based bonus available for the prediction meaningfully. Participants complete multiple versions of the game in random order, and one is randomly selected for payment.

Design details

Randomization method
Participants are randomly assigned to the following treatment conditions in a between-subject design: (i) number of rounds in the transfer game (2 vs. 3), (ii) whether calculator look-ups are free vs. carry a small per-look-up cost, and (iii) the incentive level (maximum bonus). Randomization is carried in the Qualtrics Survey by a computer. Participants then complete three game versions in randomized order, and one version is randomly selected by the software for payment.
Randomization unit
Individual-level randomization. Random assignment is conducted at the participant level (each participant is assigned to a single treatment cell in a between-subject design). There is no cluster- or group-level randomization.
Clusters
NA
Observations
Approximately 2,250 participants, Each participant completes three versions of the transfer game, yielding approximately 6,750 participant × game-version observations.
Arms
- Two-round game + Look-ups are free + Maximum bonus is $10: ~500 - Two-round game + Each look-up costs $0.01 + Maximum bonus is $10: ~500 - Three-round game + Look-ups are free + Maximum bonus is $10: ~500 - Three-round game + Each look-up costs $0.01 + Maximum bonus is $10: ~500 - Two-round game + Look-ups are free + Maximum bonus is $50: ~250
Primary outcomes
Transfers considered: Which transfers ijt are considered based on calculator look-ups (sender i, receiver j, round t), including summary measures such as the total number of transfers considered and the number of relevant transfers considered. Effects considered: Indicators for whether the participant considered each of the three effect types—direct, indirect, and (in the 3-round game) feedback—defined by whether the participant looked up all transfers along at least one path consistent with the effect. Mental model measures from look-ups: Classification of observations/participants into mental models based on the pattern of calculator look-ups, including: - Non-parametric model frequency/deviation measures (share of observations matching a given look-up pattern, or within 1–2 deviations). - Parametric (maximum-likelihood) participant classification into the mental model that best explains look-ups across the three game versions (and the implied deviation/error rate).

Cited by

No citation data available for this registration yet.

Keywords