Machine Learning Uncertainty as a Design Material: A Post-Phenomenological Inquiry
Writing this paper was simultaneously straightforward and mind-boggling. Jesse Benjamin (UTwente),Nick Merrill (UC Berkeley), James Pierce (UW Washington) and myself unfold how to consider Machine Learning Uncertainty as a Design Material through a Post-Phenomenological lens.
Let me try to summarize this philosophical/algorithmic argument here. Some definitions first. 1) Design Material: Code is as much a material as clay, yarn or glass. So, let’s consider the uncertainty of a Machine Learning algorithm as a material as well. 2) Post-Phenomenology: A strand of philosophy that studies how technology shapes the relations between humans and the world. It is useful as it provides a hands-on vocabulary to describe various relations between humans and technology (see table below).
I → (Technology–World)
I → Technology(–World)
I ↔ Technology/World
(I/Technology) ↔ World
I → (Technology → World)
(defined by Ihde and Verbeek)
Now, with this paper we use the concepts of Design Material and Post-Phenomenology to analyze and to understand how design research can engage with ML uncertainty. To do so, we look at four speculative objects from design research. The four design research objects we analyzed are: 1) Shifting Lines of Creepiness: A design-led inquiry into smart home security cameras to reflect issues of leaking surveillance in the home. 2) When BCIs Have APIs: A design fiction on Brain-Computer Interfaces for crowd workers. 3) Morse Things: A material speculation into more-than-human design of artefacts. 4) High Water Pants: A speculative design of artefacts for relating to climate change predictions. In analyzing and interpreting these four speculative objects, we propose the following concepts to understand human — machine-learning relations.
Thingly Uncertainty: ML-driven artefacts have uncertain, variable relations to their environments
Pattern Leakage: ML uncertainty can lead to patterns shaping the world they are meant to represent
Futures Creep: ML technologies texture human relations to time with uncertainty.
These three concepts can help design research to engage with the specifics of ML-driven technological mediation. We argue that these concepts offer a promising foothold for design research of ML technologies.
Design research is important for understanding and interrogating how emerging technologies shape human experience. However, design research with Machine Learning (ML) is relatively underdeveloped. Crucially, designers have not found a grasp on ML uncertainty as a design opportunity rather than an obstacle. The technical literature points to data and model uncertainties as two main properties of ML. Through post-phenomenology, we position uncertainty as one defining material attribute of ML processes which mediate human experience. To understand ML uncertainty as a design material, we investigate four design research case studies involving ML. We derive three provocative concepts: thingly uncertainty: ML-driven artefacts have uncertain, variable relations to their environments; pattern leakage: ML uncertainty can lead to patterns shaping the world they are meant to represent; and futures creep: ML technologies texture human relations to time with uncertainty. Finally, we outline design research trajectories and sketch a post-phenomenological approach to human-ML relations.
The four design research objects we analyzed:
James Pierce. 2019. Smart Home Security Cameras and Shifting Lines of Creepiness: A Design-Led Inquiry. In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems (CHI ’19). ACM, New York, NY, USA, 45:1–45:14. https://doi.org/10.1145/3290605.3300275
Ron Wakkary, Doenja Oogjes, Sabrina Hauser, Henry Lin, Cheng Cao, Leo Ma, and Tijs Duel. 2017. Morse Things: A Design Inquiry into the Gap Between Things and Us. In Proceedings of the 2017 Conference on Designing Interactive Systems (DIS ’17). ACM, NewYork, NY, USA, 503–514. https://doi.org/10.1145/3064663.3064734
Richmond Y. Wong, Nick Merrill, and John Chuang. 2018. When BCIs Have APIs: Design Fictions of Everyday Brain-Computer Interface Adoption. In Proceedings of the 2018 Designing Interactive Systems Conference (DIS ’18). ACM, New York, NY, USA, 1359–1371. https://doi.org/10.1145/3196709.3196746
Heidi R. Biggs and Audrey Desjardins. 2020. High Water Pants: Designing Embodied Environmental Speculation. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems (CHI ’20). Association for Computing Machinery, Honolulu, HI, USA, 1–13. https://doi.org/10.1145/3313831.3376429
Machine Learning Uncertainty as a Design Material at ACM CHI 2021
Jesse Josua Benjamin, Arne Berger, Nick Merrill, and James Pierce. 2021. Machine Learning Uncertainty as a Design Material: A Post-Phenomenological Inquiry. In Yokohama ’21: ACM CHI Conference on Human Factors in Computing Systems, May 8–13, 2021, Yokohama, Japan. ACM, New York, NY, USA, 14 pages. https://doi.org/10.1145/1122445.1122456