The Unbearable Lightness of Nudging: Preventing Data Leaks for Generative AI Users

Lunch Seminar in presence

Building BL26 – Room 0.19 (ground floor)
Department of Management, Economics and Industrial Engineering
Via R. Lambruschini, 4/B

Minhao Zhang
University of Bristol, UK

Abstract:

A salient risk of GenAI involves the inadvertent disclosure of sensitive or proprietary information during employees’ prompting, including supplier data, product specifications, and customer information, which may lead to privacy violations, security breaches, and reputational damage.

The study presented investigates whether behavioural nudges can effectively reduce sensitive information disclosure in procurement contexts when employees interact with GenAI systems.

To answer the research questions, this study conducted experiments using a GenAI-powered chatbot embedded in procurement tasks.The study integrated a GenAI chatbot with pre-designed nudges into procurement tasks to collect participants’ prompts and examine their disclosure behaviour regarding sensitive company information.

The results show that concurrent nudges are an effective type of nudge for preventing the disclosure of sensitive company information when users use a GenAI chatbot. Nudges appeared only when the algorithms detected potentially sensitive information during prompting. The effectiveness of concurrent nudges operates through enhanced clarity regarding sensitive information, enabling participants to better distinguish between what constitutes sensitive and non-sensitive content when explicit indications are provided.

Minhao Zhang is an Associate Professor (Reader) in Operations & Supply Chain Management from the University of Bristol Business School. The theme throughout Minhao’s current work is advancing corporate environmental management and the management of digital technologies in service operations, particularly in hospitality and healthcare. Leveraging his expertise in data analytics and supply chain management, he is co-leading REWIRE. Thanks to various cross-disciplinary research collaborations, Minhao has published high-quality papers in his field (e.g., Journal of Operations Management; Production and Operations Management; International Journal of Operations & Production Management; Journal of Association for Information Systems) and earned top-cited and most-downloaded article awards from both the British Academy of Management and the European Academy of Management. He is the Associate Editor of IJOPM.