Interaction Density as a Behavioural Signature of Exhibit Type: A Minimal-Log Study from a Two-Venue Science Experience Centre

Science centres often struggle to quantify visitor engagement because traditional methods like manual observation or invasive sensor deployment are costly and difficult to scale. This study investigates whether simple, default interaction logs—specifically session start time, end time, and total press count—can serve as a reliable, privacy-preserving proxy for understanding visitor behaviour across different types of interactive exhibits.

The researchers analysed 2,816 visitor sessions across eight exhibits at two science centre venues in Bengaluru, India. They derived a metric called interaction density (presses per second) and tested three hypotheses comparing this density, session duration (dwell time), and press counts between two exhibit categories: games (goal-directed and progressively difficult) and quizzes (structured, non-adaptive questions). The study used Mann-Whitney U tests for group comparisons, logistic regression for classification, and OLS regression to evaluate the predictive power of these metrics.

Interaction density successfully distinguishes between games and quizzes with a cross-validated AUC of 0.778, proving that minimal logs can capture meaningful behavioural differences. However, the data challenged the researchers' initial assumptions. While games were indeed more intense, they also resulted in longer dwell times, reversing the intuitive trade-off between intensity and duration. This was largely driven by exhibits with escalating difficulty that created open-ended re-engagement loops. Furthermore, while density is a useful classifier, raw press count is a significantly stronger predictor of dwell time, suggesting that density and press count should be used as complementary metrics rather than substitutes.

This research demonstrates that science centres do not need expensive, invasive sensing hardware to conduct rigorous behavioural research. By leveraging existing touch-interface logs, institutions can generate falsifiable insights into how visitors interact with different exhibit designs. This low-cost, scalable methodology allows for continuous, data-driven exhibit evaluation, helping designers identify anomalies and refine visitor experiences without compromising privacy or incurring high operational costs.