Point of contact
Jules Françoise, LISN
Machine Learning (ML) is a powerful tool for building applications that perform complex tasks using computational models estimated from data. While it has become ubiquitous in a wide array of softwares and services, it often remains conceived as a black box that works autonomously on passively collected data. Yet, this viewpoint hides considerable human work of tuning the algorithms, gathering the data, and even deciding what should be modeled in the first place. A growing community at the intersection of Machine Learning and Human Computer Interaction investigates human-centered perspectives on ML that explicitly recognise this human work, reframe machine learning workflows based on situated human working practices, and explore the co-adaptation of humans and systems.
In a traditional machine learning workflow, practitioners collect data, select or engineer features to represent the data, choose a learning algorithm and fine-tune its parameters, and finally assess the quality of the resulting model. This workflow results in long iterations, which limit the user’s ability to interact with the model and affect its results. Interactive machine learning is a research topic at the intersection of ML and HCI, where learning cycles involve more rapid, focused, and incremental model updates than in the traditional machine learning process. Such an approach can empower users to create ML-based systems for their own needs and purposes. However, enabling effective end-user interaction with interactive machine learning introduces new challenges that require a better understanding of end-user capabilities, behaviors, and needs. Research questions in interactive and user-centered machine learning include:
- Supporting the design of interactions by demonstration using machine learning, which requires developing new models and interaction techniques to help novice users interact efficiently with ML and AI algorithms.
- Understanding and facilitating end-user interaction with ML and AI systems, in particular for people with limited to no expertise in ML. This is critical in expert domains, for example in the medical field where ML-based systems are developed to assist clinicians, and requires developing visualisations, explanations, or interaction techniques to increase trust and facilitate user understanding of ML predictions.
- Supporting the workflow of machine learning practitioners through novel visualisations and interactions to help them build more robust models, assess them efficiently in real-life scenarios, and reduce their inherent biases.
Researchers involved or interested
A few references
- Amershi, S., Cakmak, M., Knox, W. B., & Kulesza, T. (2014). Power to the people: The role of humans in interactive machine learning. Ai Magazine, 35(4), 105-120.
- Dudley, J. J., & Kristensson, P. O. (2018). A review of user interface design for interactive machine learning. ACM Transactions on Interactive Intelligent Systems (TiiS), 8(2), 1-37.
- Fiebrink, R., Cook, P. R., & Trueman, D. (2011, May). Human model evaluation in interactive supervised learning. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (pp. 147-156).
- Françoise, J., & Bevilacqua, F. (2018). Motion-sound mapping through interaction: An approach to user-centered design of auditory feedback using machine learning. ACM Transactions on Interactive Intelligent Systems (TiiS), 8(2), 1-30.
- Boukhelifa, N., Bezerianos, A., & Lutton, E. (2018). Evaluation of interactive machine learning systems. In Human and Machine Learning (pp. 341-360). Springer, Cham.
- Caramiaux, B., & Tanaka, A. (2013). Machine learning of musical gestures: Principles and review. In Proceedings of the International Conference on New Interfaces for Musical Expression (NIME) (pp. 513-518). Graduate School of Culture Technology, KAIST.