This research investigates the helpfulness of AI-generated review summaries on e-business platforms. Drawing on the cognitive-affective model of decision-making, we examine how cognitive and affective content influence the perceived helpfulness of AI review summaries. Study 1 analyzes 1274 AI review summaries from Newegg.com and shows that cognitive content increases perceived helpfulness, whereas positive and negative affective content reduce perceived helpfulness. The positive effect of cognitive content weakens as product price increases. Study 2, an online experiment, replicates the preference for content with stronger cognitive emphasis over content with stronger affective emphasis in AI review summaries and demonstrates that this preference is associated with the AI-generated nature of the content. This research advances understanding of user responses to AI-generated content and provides insights for improving generative AI applications in online content curation.
Contemporary societal challenges heighten the importance of designing information systems to contribute to improving societal outcomes. While design science research (DSR) has progressively expanded from business settings to broader societal domains, such as healthcare, environmental sustainability, social inclusion, and public governance, the notion of societal outcome is often confused with output and impact. This essay synthesizes insights from the ten papers in this special issue toward improving societal outcomes. Building on this synthesis, we propose an artifact-enabled societal outcome framework. The framework frames societal outcomes as not being achievable through stakeholder engagement alone, but through a rigorous scientific process that combines curiosity in problem framing, timely intervention, artifact validity, and evidence that artifact outputs translate into societal outcomes.
Predicting the behavior of newly acquired users is a central challenge for digital platforms because direct behavioral histories are sparse at entry. We examine whether relational information revealed by a user’s first transaction can improve decision support for cold-start prioritization in peer-to-peer (P2P) platforms. Using data on approximately 198,000 Venmo users, we compare models based on attributes of the focal user, the first counterparty, and the counterparty’s pre-existing community. We study two managerially relevant outcomes: near-term inactivity risk and future high-engagement-user identification. At entry, adding relational information increases top-decile lift from 1.11 to 1.20 for inactivity risk and from 1.22 to 1.53 for high-engagement identification. Community information also provides additional value beyond the first counterparty, particularly for identifying future high-engagement users. Decomposing the community signal shows that recent transactional activity is more informative than network structure alone. As an interpretable complement, a selection-adjusted analysis shows that entering through a more cohesive community is associated with a 5.0 to 6.5 percentage-point higher probability of becoming a high-engagement user. These findings show how information embedded in the community revealed by a user’s first transaction can support early prioritization when individual behavioral information is limited.
