Artificial intelligence (AI) is increasingly used in organizations to automate tasks and processes. However, many organizations struggle with AI models taking decisions that deviate from their expectations. Usually this happens not due to missing model capabilities, but due to the model not being aligned to its organizational context: It is missing the tacit knowledge that employees have in order to decide in alignment with expectations.
To understand this better, we are looking for students willing to investigate one type of decision that occurs in organizations of your choice. For example, a thesis could look at insurance claim decisions and gather qualitative or/and quantitative data about how humans take decisions on edge cases. Qualitative data could be gathered in interviews with decision- makers to identify all cues influencing the decision and their subjective weighting. Quantitative data could be gathered from organizations willing to share (anonymized) datasets for past decisions (data that the decision was based on + the human decision itself) or by designing and implementing a pilot experiment testing contextualized decision-making.
As this is closely related to my own research, I will be able to provide close mentoring and theoretical framing for the thesis project.
If you are interested, please send an email to niklas.weller@unisg.ch. I am happy to set up a call/meeting to discuss further details or the fit to your own thesis ideas.