Too Narrow to Help? Unveiling How Recommendation Agents’ Specialization Impacts User Choices

Abstract

On many online platforms, professional human recommenders have largely been replaced by recommendation agents (RAs): algorithms that can—at lower cost and higher speed—incorporate users’ explicit and tacit preferences into customized search results that help with the purchase decision process. RAs are often built around understanding users’ past preferences to make accurate recommendations that generally reinforce said preferences. This approach offers several advantages but also limits consumers’ ability to consider options outside of their past interests—the so-called specialization issue. The present research hypothesizes that a specialized RA (vs. a generalized preference-weighted RA) reduces users’ willingness to accept the recommendation. This effect is sequentially mediated by users’ perceived breadth of knowledge, perceived control over the choice process, and perceived reciprocity with the RA. To test these hypotheses, the authors programmed a functioning RA and implemented it in three experimental studies involving 705 online participants. Results confirm the hypotheses suggesting that users do sometimes want RAs to help them expand on, rather than merely reaffirm, their existing preferences, particularly when their product expertise is relatively low. Theoretical and managerial implications as well as avenues for future research are discussed.