The Threatening Effects of Help from Agentic Information Systems and their Implications for User-System Collaboration (Harmful Help from Information Systems)

• Applicant: Dr. Martin Adam
• Subject Area: Operations Management and Computer Science for Business Administration ; Data Management, Data-Intensive Systems, Computer Science Methods in Business Informatics
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 522190413

The goal of the research project is to examine the nature and effects of help invocation types (HITs) from agentic information systems (IS) and their implications for user-IS collaboration. Although some previous studies have differentiated between user-invoked and IS-invoked help (e.g., user- or IS-initiated provision of advice or execution of tasks), these studies are mainly conceptual, and still little understanding exists of how users react to different HITs and particularly to anticipatory help (i.e., the IS invokes the offering of help) in contrast to reactive help (i.e., the user invokes help from the IS). Without knowing how, why, and when users react to anticipatory (vs. reactive) help, providers of agentic IS may employ anticipatory help expecting the same or even better user reactions as when they design for reactive help, with potentially detrimental consequences for the user-IS collaboration. Indeed, even though anticipatory help may be objectively useful for many user-IS collaborations, users may perceive self-threats to their identity (e.g., competence, status) by IS initiating interactions to better solve a task, so that users may not only disregard the offered anticipatory help but even stop interacting with the IS, ultimately turning a well-meant design of agentic IS into unintended sabotage of the user-IS collaboration. Against this backdrop, we aim to develop with this research project validated conceptual definitions and taxonomy for HITs based on a sound theoretical grounding and comprehensive empirical tests. Furthermore, we plan to validate reactive and anticipatory help as the most prevailing HITs in a well-advanced theoretical model (i.e., Self-Affirmation Theory) based on a series of randomized online experiments. We aim to mainly investigate the short-term effects of HITs on (subjective) user perceptions, that is particularly perceived self-threat to a user's identity, as well as boundary conditions that amplify or mitigate these threatening effects. Lastly, we aim to complement and corroborate these findings by exploring (long-term) user-IS collaboration outcomes from an instrumental (e.g., accepting help, performance) and humanistic perspective (e.g., satisfaction, perceived exploitation). For this purpose, we intend to cooperate with two companies in employing different HITs in large-scale field experiments. With the research project, we want to contribute to the IS help and user-IS collaboration literature by theoretically providing a more nuanced understanding of HITs and their implications and designs, which becomes increasingly important with the spread of more agentic IS. Given that the project operates at the interface between IS, human-computer interaction, and related fields, we expect that our results will also influence an audience that extends further beyond the IS community.

• DFG Programme: Research Grants