Exploring the economics of conversational search sessions

Abstract

Purpose

Turn-taking is beneficial to conversational search success, but the increase in turns and time can also increase the cognitive load of the user. Therefore, in this research paper, the authors view conversational search sessions through the lens of economic theory and use the economic models of search to analyze the various costs and benefits of information-seeking interactions.

Design/methodology/approach

First, the authors built a cost-benefit model for conversational search sessions by defining action types and performing an intellectual mapping of actual sessions into sequences of these actions (using thematic analyses). The authors used the hypothesized cost and benefit actions (obtained from the user-system dialogs), along with the number of turns, utterances and time-related parameters, to propose the mathematical model. Next, the authors tested the model empirically by comparing the model scores to the user satisfaction and task success scores (collected through questionnaires). By representing each session as a bag of actions, the authors developed linear regression models to predict task success and user satisfaction.

Findings

Through feature analysis and significance testing, the authors identify the different parameters that contribute significantly to user satisfaction and task success scores. Error analysis shows that the model predicts task success and user satisfaction reasonably well, with the average prediction error being 0.5 for both (on a 5-point scale).

Originality/value

The authors' research is an initial step toward building a mathematical model for predicting user satisfaction and task success in conversational search sessions.

Citation

Ghosh, S., Gogoi, J. and Chua, K. (2023), "Exploring the economics of conversational search sessions", Aslib Journal of Information Management, Vol. ahead-of-print No. ahead-of-print. https://doi.org/10.1108/AJIM-08-2022-0368