RIROSE: Rational Information Retrieval in Online Search Environments

Debora Di Caprio,Francisco J. Santos-Arteaga

doi:10.1016/j.simpa.2022.100248

Debora Di Caprio, Francisco J. Santos-Arteaga

Open Access

https://doi.org/10.1016/j.simpa.2022.100248

Copy DOI

Abstract

<h2>Abstract</h2> We design a benchmark algorithm that mimics the sequential behavior of users when retrieving information from the set of alternatives provided by an engine within the first page of online search results. The benchmark defined by the algorithm is designed to evaluate deviations from the rational retrieval strategies determined by the subjective preferences and beliefs of users. The algorithm accounts for the 2047 nodes composing the binary decision tree defined by the ten alternatives ranked within the first page of results. The flexibility of the algorithm allows to incorporate modifications accounting for search frictions and different degrees of impatience on the side of users, as well as testing the categorization capacities of machine learning techniques.

Highlights

We design a benchmark algorithm that mimics the sequential behavior of users when retrieving information from the set of alternatives provided by an engine within the first page of online search results
The benchmark algorithm defines an information retrieval framework where each decision maker (DM) sets out to evaluate the ten alternatives composing the initial page of results provided by a search engine
This latter value is subjectively defined by the DM based on his beliefs regarding the potential realizations that may be observed from the characteristics composing the different alternatives

Summary

Support email for questions

The main intuition on which the algorithms described in this paper are built is the design of an information retrieval process that is as close as possible to the behavior of a standard rational decision maker (DM) formalized in the decision theoretical literature while accounting for the sequential evaluation structures common to the main empirical studies. The benchmark algorithm defines an information retrieval framework where each DM sets out to evaluate the ten alternatives composing the initial page of results provided by a search engine. The DM follows a sequential process determined by the observations retrieved from the alternatives displayed, which condition his subsequent retrieval behavior

Code Ocean

Pages clicked

Declaration of competing interest