Engagement in proactive recommendations

Laurens Rook,Markus Zanker,Adem Sabic

doi:10.1007/s10844-018-0529-0

Abstract

The present research explored to what extent user engagement in proactive recommendation scenarios is influenced by the accuracy of recommendations, concerns with information privacy, and trait personality. We hypothesized that people’s self-reported information privacy concerns would matter more when they received accurate (vs. inaccurate) proactive recommendations, because these pieces of advice would seem fair to them. We further hypothesized that this would particularly be the case for people high on the social personality trait Extraversion, who are by inclination prone to behaving in a more socially engaging manner. We put this to the test in a controlled experiment, in which users received manipulated proactive recommendations of high or low accuracy on their smartphone. Results indicated that information privacy concerns positively influenced a user’s engagement with proactive recommendations. Recommendation accuracy influenced user engagement in interaction with information privacy concerns and personality traits. Implications for the design of human-computer interaction for recommender systems are addressed.

Highlights

Recommender systems (RS) are automated decision support tools designed to provide custom-made advice on items to facilitate people’s navigation in large product or information spaces
The analysis yielded a significant main effect of low versus high recommendation accuracy, β = .60, t (126) = 2.67, p < .01, indicating that participants had experienced the quality of the proactive recommendations they had received in congruence with their experimental condition
The other descriptives in the table derive from a list of covariate items that are more often collected in studies on information privacy concerns (Malhotra et al 2004; Smith et al 1996)

Summary

Introduction

Recommender systems (RS) are automated decision support tools designed to provide custom-made advice on items to facilitate people’s navigation in large product or information spaces. RS offer personalized suggestions based on presumed needs and preferences of a user and other people’s behavior They help people overcome information overload, either by providing recommendations on request, or by delivering them proactively (Jannach et al 2010; Ricci et al 2015). A potential side-effect of accuracy as a quality metric is that the RS may recommend items too similar to what the user liked before Such recommendations may no longer satisfy the user (Konstan and Riedl 2012; McNee et al 2006). Accuracy as a quality metric, remains a necessary condition; if the RS is not accurate, the user will not trust it (Jannach et al 2016) This element of trustworthiness potentially turns accurate RS into powerful tools to deal with people’s concerns with information privacy during navigation in large product or information spaces. Results of our experimental study show that all of the three factors positively influenced user engagement

Methods

Results

Conclusion