Prof. Hartmann from Technical University Munich (TUM) has recently published a paper in the FT50-ranked Journal of Consumer Psychology, co-authored by Anouk Bergner and Christian Hildebrand from the Institute of Behavioral Science and Technology at the University of St.Gallen (IBT-HSG). The paper introduces "MindMiner," a language model designed to predict mind perception in smart objects derived from consumer-generated text.
The team employed a variety of text mining techniques and machine learning methods, such as keyword-assisted topic models (keyATM) and local interpretable model-agnostic explanations (LIME), to discern a broad set of linguistic markers for mind perception. This comprehensive analysis is based on the scrutiny of over 20,000 real customer reviews.
The study found marked heterogeneity in how consumers relate to their smart objects, like Amazon's Alexa or conversational agents such as ChatGPT. Some consumers perceive these devices as merely functional tools for specific tasks, while others view them as helpful partners with which they engage in a communal relationship. This diversity in relationships translates into substantial differences in smart-object usage intensity, or the number of tasks consumers assign to their smart assistants.
In line with encouraging open-source models for academic and research purposes, the team has made MindMiner publicly available on Hugging Face. They demonstrate in their paper how MindMiner can be fine-tuned with minimal training data to predict empathy expressions from text, among other tasks.
This research is part of a special issue on text analytics in JCP co-edited by Grant Packard, Sarah Moore, and Jonah Berger. It aligns with TUM School of Management's interdisciplinary mission, bridging management and technology.
For a more detailed view of the team's methods and findings, the full paper is available and will shortly be provided as open access.
The complete citation for the publication is: Hartmann, J., Bergner, A., & Hildebrand, C. (2023). MindMiner: Uncovering Linguistic Markers of Mind Perception as a New Lens to Understand Consumer-Smart Object Relationships. JCP. Forthcoming.