Effective Human-AI Collaboration in ML Operations

Thesis (BA/MA)

Advisor(s): Michael Weber (mic.weber@tum.de)

Context

Many of today’s systems labelled “Artificial Intelligence” (AI) rely on machine learning (ML) techniques. ML enables machines to learn and improve from experience, such as historical datasets (Jordan & Mitchell, 2015). ML offers manifold opportunities to transform businesses and society by enabling new products and services (Weber et al., 2022a), the automation of tasks, and the augmentation of human capabilities (Raisch & Krakowski, 2021). For example, ML can augment radiologists in their diagnoses to make healthcare services more efficient and reliable for the greater good (e.g., Lebovitz et al., 2021).

However, currently much of the anticipated value remains unrealized, as many organizations struggle to successfully operate ML systems (Benbya et al. 2020). One important challenge is to maintain an ML system’s performance throughout its lifecycle, as various, oftentimes dynamic environmental factors can have significant influence on the performance (e.g., changing consumer buying habits) (Paleyes et al., 2022). Most of today’s ML systems are typically limited in their learning abilities (Balasubramanian et al., 2022), rendering these ML systems unable to autonomously respond to such changes in their environment. Therefore, humans are often required to work for the AI, e.g., by monitoring for changes and adjusting the ML systems to sustain performance (Lyytinen et al., 2021).

Current literature highlights the need for a strong ML operations capability to create value from ML systems (Shollo et al., 2022, Weber et al., 2022b). However, it so far remains unclear how humans engage with ML systems and their operating environments to successfully build such a ML operations capability. Some studies have already provided early insights on human expert’s roles in auditing and adjusting an ML system (Gronsund & Aanestad, 2020), making sense of ML predictions (Lebovitz et al., 2021), or controlling the boundaries of an ML system (Asatiani et al., 2021). Thus, the goal of this thesis is to provide a more detailed understanding of how humans engage with ML systems and their operating environment to ensure sustained performance.

Potential research questions include (but are not limited to):

  • What are the roles of human experts and ML engineers in operating ML systems?
  • What are practices for monitoring and updating ML systems?
  • Which factors make a ML system’s environment more or less dynamic?
  • What are potential practices for creating a more stable operating environment?
  • How does ML operation differ for different kinds of ML projects?

Task(s)

  • Review literature in the respective field
  • Collect and analyze empirical data on real-world ML projects (e.g., expert interviews)
  • Derive practical guidelines for ML operations
  • Discuss implications research

Requirements

  • High degree of autonomy and individual responsibility
  • Good communicative skills
  • Basic understanding in machine learning is beneficial

Further Information

The thesis can be written in English or German. The topic can also be adapted to your interests. If you have further questions, please do not hesitate to contact me directly. Please send your application including our application form, a current transcript of records, and your CV to mic.weber@tum.de. Please note that we can only consider applications with complete documents. 

References

Asatiani, A., Malo, P., Nagbøl, P. R., Penttinen, E., Rinta-Kahila, T., & Salovaara, A. (2021). Sociotechnical Envelopment of Artificial Intelligence: An Approach to Organizational Deployment of Inscrutable Artificial Intelligence Systems. Journal of the Association for Information Systems, 22(2), 325-352.

Balasubramanian, N., Ye, Y., & Xu, M. (2022). Substituting human decision-making with machine learning: Implications for organizational learning. Academy of Management Review47(3), 448-465.

Benbya, H., Davenport, T. H., & Pachidi, S. (2020). Artificial Intelligence in Organizations: Current State and Future Opportunities. MIS Quarterly Executive, 19(4), ix-xxi.

Grønsund, T., & Aanestad, M. (2020). Augmenting the algorithm: Emerging human-in-the-loop work configurations. The Journal of Strategic Information Systems29(2), 101614.

Jordan, M. I., & Mitchell, T. M. (2015). Machine learning: Trends, perspectives, and prospects. Science, 349(6245), 255-260.

Lebovitz, S., Levina, N., & Lifshitz-Assaf, H. (2021). Is AI Ground Truth Really True? The Dangers of Training and Evaluating AI Tools Based on Experts’ Know-What. Mis Quarterly, 45(3), 1501-1525.

Lyytinen, K., Nickerson, J. V., & King, J. L. (2021). Metahuman systems= humans+ machines that learn. Journal of Information Technology, 36(4), 427-445.

Paleyes, A., Urma, R. G., & Lawrence, N. D. (2022). Challenges in deploying machine learning: a survey of case studies. ACM Computing Surveys, 55(6), 1-29.

Raisch, S., & Krakowski, S. (2021). Artificial intelligence and management: The automation–augmentation paradox. Academy of Management Review, 46(1), 192-210.

Shollo, A., Hopf, K., Thiess, T., & Müller, O. (2022). Shifting ML value creation mechanisms: A process model of ML value creation. The Journal of Strategic Information Systems31(3), 101734.

Weber, M., Beutter, M., Weking, J., Böhm, M., & Krcmar, H. (2022a). AI Startup Business Models. Business & Information Systems Engineering64(1), 91-109.

Weber, M., Engert, M., Schaffer, N., Weking, J., & Krcmar, H. (2022b). Organizational capabilities for ai implementation—coping with inscrutability and data dependency in ai. Information Systems Frontiers, 1-21.