Technical Understanding from IML Hands-on Experience: A Study through a Public Event for Science Museum Visitors

TL;DR

Non-experts can develop basic technical comprehension of ML through hands-on IML in public settings, as shown by a science-museum event using an interactive sound-classification system with feature-map visualization. The study collects rich data from logs, questionnaires, quizzes, and interviews to assess comprehension, revealing that many participants gain concepts and confidence without explicit instruction, though prior AI experience can bias interpretation and usage. Findings highlight both the potential and limitations of IML as an educational tool for broad audiences, and they provide concrete design implications for future IML interfaces and events aimed at improving ML literacy. The work suggests that thoughtful task design, collaborative discussion, and interpretable visualizations are key to maximizing understanding in non-expert populations across diverse ML tasks and modalities.

Abstract

While AI technology is becoming increasingly prevalent in our daily lives, the comprehension of machine learning (ML) among non-experts remains limited. Interactive machine learning (IML) has the potential to serve as a tool for end users, but many existing IML systems are designed for users with a certain level of expertise. Consequently, it remains unclear whether IML experiences can enhance the comprehension of ordinary users. In this study, we conducted a public event using an IML system to assess whether participants could gain technical comprehension through hands-on IML experiences. We implemented an interactive sound classification system featuring visualization of internal feature representation and invited visitors at a science museum to freely interact with it. By analyzing user behavior and questionnaire responses, we discuss the potential and limitations of IML systems as a tool for promoting technical comprehension among non-experts.

Figures (9)

• Figure 1: This study investigates the effect of interactive machine learning (IML) on end users' technical comprehension. We hosted a hands-on event in a science museum to invite public visitors.
• Figure 2: The overview of our IML system (translated from Japanese). The system has three major components, allowing users to create sound classification models.
• Figure 3: The interaction flow of our system (translated). In the category-driven flow, users first define target categories and then search for appropriate sound samples on the map or directly record the sound. In the sample-driven flow, users first record a sound on the feature map and then register sounds to the categories. They then train the classification model with the training data and validate its performance in both cases.
• Figure 4: Illustrations of the worksheet used at the event (translated from Japanese).
• Figure 5: Overall trends in questionnaire responses.