A Human-Centered Approach to Designing Teachable Systems
Proceedings of the 2019 CHI Workshop - Where is the Human? Bridging the Gap Between AI and HCI
2019
Abstract
Machine Learning technologies will eventually enable an unprecedented level of empowerment for end users–letting them adapt and personalize the behavior of AI systems to their own unique situations, needs, and desires through naturalistic interactions. Similar to how people can improve and extend one another through teaching, machine learning technologies promise to let users modify the behavior of systems by teaching rather than programming. However, there are many challenges faced when designing such teachable systems that have heretofore impeded their development. This paper aims to highlight two of these challenges. First, the development of learning systems has largely been restricted to technical fields, such as machine learning or cybernetics, due to the complexity of making such systems even work. Thus, there has been a lack of opportunities to provide an HCI perspective on how interactions with such systems should be designed. To address this challenge, we introduce a framework for conceptualizing teachable systems that distinguishes the interactive component from the learning component of these systems, to better enable the application of the HCI perspective on the design of the interaction between the user and system. Second, there is also a general lack of HCI methods for prototyping teachable systems, specifically when it comes to designing systems with learning capabilities that do not currently exist. To overcome this challenge, we presents a novel variation of the classic Wizard-of-Oz prototyping paradigm that specifically supports the rapid prototyping of interactive learning systems.
BibTeX
@inproceedings{maclellan-chi-2019-ws,
title = {A Human-Centered Approach to Designing Teachable Systems},
author = {MacLellan, Christopher J. and Harpstead, Erik and Sheline, Rob},
booktitle = {Proceedings of the 2019 CHI Workshop - Where is the Human? Bridging the Gap Between AI and HCI},
year = {2019},
}