Interacting meaningfully with machine learning systems: Three experiments

Simone Stumpf,Vidya Rajaram,Lida Li,Weng-Keen Wong,Margaret Burnett,Thomas Dietterich,Erin Sullivan,Jonathan Herlocker

doi:10.1016/j.ijhcs.2009.03.004

Simone Stumpf, Vidya Rajaram + Show 6 more

Open Access

https://doi.org/10.1016/j.ijhcs.2009.03.004

Copy DOI

Abstract

Although machine learning is becoming commonly used in today's software, there has been little research into how end users might interact with machine learning systems, beyond communicating simple “right/wrong” judgments. If the users themselves could work hand-in-hand with machine learning systems, the users’ understanding and trust of the system could improve and the accuracy of learning systems could be improved as well. We conducted three experiments to understand the potential for rich interactions between users and machine learning systems. The first experiment was a think-aloud study that investigated users’ willingness to interact with machine learning reasoning, and what kinds of feedback users might give to machine learning systems. We then investigated the viability of introducing such feedback into machine learning systems, specifically, how to incorporate some of these types of user feedback into machine learning systems, and what their impact was on the accuracy of the system. Taken together, the results of our experiments show that supporting rich interactions between users and machine learning systems is feasible for both user and machine. This shows the potential of rich human–computer collaboration via on-the-spot interactions as a promising direction for machine learning systems and users to collaboratively share intelligence.

Highlights

A new style of human-computer interaction is emerging, in which some reasoning and intelligence reside in the computer itself
Rule-based explanations were consistently understandable to more than half the participants and, at least for this group of participants, seemed to "win" over the other two paradigms, note that about one-third of the participants preferred one of the other explanation paradigms. This implies that machine learning systems may need to support multiple explanation paradigms in order to effectively reach all of their users
Instead of drawing attention to words with negative weights that are present in the email, the explanation could make use of the strongest negative weights associated with words that are absent from emails, since their absence increases the confidence of the learning algorithm

Summary

Introduction

A new style of human-computer interaction is emerging, in which some reasoning and intelligence reside in the computer itself. These intelligent systems and user interfaces attempt to adapt to their users’ needs, to incorporate knowledge of the individual’s preferences, and to assist in making appropriate decisions based on the user's data history. These approaches use artificial intelligence to support the system’s part of the reasoning. One increasingly common approach being brought to intelligent systems and user interfaces is machine learning, in which the system learns new behaviors by examining usage data. Interactive email spam filters, such as in Apple’s Mail system, are prime examples

Methods

Results

Discussion

Conclusion