Please fill out this short user survey of only 3 questions in order to help us improve the site. We appreciate your feedback!
The Role of Memory in Beliefs Formation
Last registered on August 01, 2020


Trial Information
General Information
The Role of Memory in Beliefs Formation
Initial registration date
August 09, 2019
Last updated
August 01, 2020 5:42 AM EDT
Primary Investigator
Other Primary Investigator(s)
PI Affiliation
Microsoft Research and NBER
PI Affiliation
University of Michigan
Additional Trial Information
Start date
End date
Secondary IDs
This experiment studies how people memory limitations affect the process of beliefs formations.
External Link(s)
Registration Citation
Mobius, Markus, Tanya Rosenblat and Pierre-Luc Vautrey. 2020. "The Role of Memory in Beliefs Formation." AEA RCT Registry. August 01. https://doi.org/10.1257/rct.4526-1.1.
Former Citation
Mobius, Markus et al. 2020. "The Role of Memory in Beliefs Formation." AEA RCT Registry. August 01. http://www.socialscienceregistry.org/trials/4526/history/73442.
Experimental Details
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
Double counting. Effect of time decay (through distraction) on double counting.
Primary Outcomes (explanation)
The outcomes are constructed using logit belief updates regressions which are described formally in the PDF attached.
Secondary Outcomes
Secondary Outcomes (end points)
Can double counting be explained only by limited attention and imperfect recognition?
Secondary Outcomes (explanation)
This is based on theoretical bounds on double counting that we derive in the attached PDF.
Experimental Design
Experimental Design
The experiment consists in having participant read fictional news about a fictional company and reporting their beliefs about the state of the world, which is an attribute of the company influencing the production of news. Some news will organically appear repeatedly while others will not and we focus on understand how memory and attention limitations shape over-reaction to repeated news.

Experimental Design Details
The experiment consists of three stages.

During the news generation and presentation
stage, we generate a valence-neutral set of eight facts stylized
as newspaper snippets drawn without replacement from a base set of
twelve facts about a fictional company. Each fact exists in a positive and a negative version.
Each fact is then assigned
a valence of positive or negative to define the subject-specific
set of eight facts. Each subject will be shown and asked to carefully
read two independent news pages, where each news page consists of
five facts drawn without replacement from the subject-specific set
of facts. The content of each fact remains the same on each news
page whereas the exact wording of each fact differs slightly from
page to page. Subjects will then be asked about their belief of the quality of the company's management conditional on the newspaper snippets they will have read in the belief elicitation stage. In the recognition stage, subjects will be
shown twelve facts: the eight facts in the subject-specific set and
the four other facts from the base set, and will be tasked with recognizing
whether or not they had read each snippet. This will be followed by a facts free recall task where subjects are asked to recall as many facts as they can, as accurately as possible. Finally, all the subjects will complete a words free recall task at the end (unrelated to the other sections) in order to get an independent measure of their recall performance. The experiment has two treatment arms. The first
treatment arm is the decay treatment: subjects complete a
simple three-minute long incentivized task; participants
in the no-decay control group complete the filler task at the
end of the experiment whereas the decay-inducing treatment group complete
the filler task in between the news generation and belief
elicitation stages. Screenshots of
the experiment from the subjects' perspective are available in the attached PDF.
Randomization Method
Randomization done by a computer.
Randomization Unit
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
700 individuals
Sample size: planned number of observations
700 invidivuals
Sample size (or number of clusters) by treatment arms
350 in Decay-Inducing treatment, 350 in No Decay treatment
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
We used a bootstrap approach to power calculation using pilot data. See the attached PDF.
IRB Name
MIT Committee on the Use of Humans as Experimental Subjects (COUHES)
IRB Approval Date
IRB Approval Number
Analysis Plan

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Post Trial Information
Study Withdrawal
Is the intervention completed?
Is data collection complete?
Data Publication
Data Publication
Is public data available?
Program Files
Program Files
Reports, Papers & Other Materials
Relevant Paper(s)