Which interpretation to follow? How do people infer what causes what?

How to cite

Ambuehl, Sandro; Thysen, Heidi Christina. 2023. "Which interpretation to follow? How do people infer what causes what?." AEA RCT Registry. AEARCTR-0011336. https://doi.org/10.1257/rct.11336-1.0

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

Sandro Ambuehl* , Heidi Christina Thysen

* Primary investigator

Registry ID
AEARCTR-0011336
DOI
https://doi.org/10.1257/rct.11336-1.0
First registered
Last updated
Status
Withdrawn
Trial dates
2023-04-24 – 2023-05-13
Countries
Switzerland ()
JEL codes
C91, D01, D83
Registry record
http://www.socialscienceregistry.org/trials/11336

Abstract

A central factor when choosing an action is its causal effect on outcomes we care about. Yet, for many decisions, only observational data are available. While experts often provide causal interpretations and recommendations for action, they frequently disagree. We use a laboratory experiment to study how humans choose in such settings. Our results inform the nascent literature on narrative competition and on model persuasion.

Intervention

In each decisions subjects are faced with a choice between either two or three advisors. By choosing an advisor the subject (by design) also chooses the action recommended by the chosen advisor. The actions are costly and maps into the subjects bonus payment according to a data generating process (DGP) that is not directly known to the subjects. To aid their the decision the subjects can choose to inspect graphical representations of the data.

Experimental design

In each decisions subjects are faced with a choice between either two or three advisors. By choosing an advisor the subject (by design) also chooses the action recommended by the chosen advisor. The actions are costly and maps into the subjects bonus payment according to a data generating process (DGP) that is not directly known to the subjects. To aid their the decision the subjects can choose to inspect graphical representations of the data.

Design details

Randomization method
Computer
Randomization unit
Each subject makes multiple decision, some randomization is at the decision level, other is at the subject level
Clusters
700 student subjects
Observations
700 student subjects
Arms
700 student subjects, 29 or 30 decisions for each (depending on treatment)
Primary outcomes
Choice of advice
Secondary outcomes
Interaction data

Cited by

No citation data available for this registration yet.

Keywords