The Emergence of Interactive Behavior

Abstract

One reason that human interaction with technology is difficult to understand is because the way in which people perform interactive tasks is highly adaptive. One such interactive task is menu search. In the current article we test the hypothesis that menu search is rationally adapted to (1) the ecological structure of interaction, (2) cognitive and perceptual limits, and (3) the goal to maximise the trade-off between speed and accuracy. Unlike in previous models, no assumptions are made about the strategies available to or adopted by users, rather the menu search problem is specified as a reinforcement learning problem and behaviour emerges by finding the optimal policy. The model is tested against existing empirical findings concerning the effect of menu organisation and menu length. The model predicts the effect of these variables on task completion time and eye movements. The discussion considers the pros and cons of the modelling approach relative to other well-known modelling approaches.

Effects of Frequency Distribution on Linear Menu Performance

“…There is a continuous effort to encapsulate HCI knowledge by building models of users' behavior. Recently, models of menu performance [2,5,6] do not only predict average menu selection time, but also item selection time. Our results indicate that we should take frequency distribution into account when modeling menu performance so that the distribution is correctly parameterized and the investigation is ecologically valid.…”

“…While uniform distribution does not reflect menu usage in real worlds [8,9,10,12], it has been used in many menu studies [2,4,5,16,20] because it is easy to experimentally control. In contrast, Zipfian distribution is more difficult to control in spite of better modelling item frequency distribution in common applications [6,8,9].…”

“…As a consequence many scientific efforts have been devoted to better understand user's behavior in linear menus [2,5,6] and how the menu features impact performance. For instance, several empirical studies described how menu length, menu organization or target position affect performance [4,21].…”

“…While it is well known [6,8,9], the consequences of the distribution for behavior have not been systematically investigated. Indeed, many previous studies rely either on a Uniform distribution [2,4,5,16,20] when comparing selection time, or a variant of a Zipfian distribution [1,11] when focusing on the average menu selection time. However, it is unclear if these observations hold when given a different, and more ecologically plausible, frequency distribution.…”

Parameter Inference for Computational Cognitive Models with Approximate Bayesian Computation

“…The proposed methods are promising for parameter estimation in ACT-R, where simulations are often computationally intensive and do not permit the computation of a likelihood. The second case is a model that predicts eye movements and task completion time when searching for a target item from a visually displayed computer drop-down menu (Chen, Bailly, Brumby, Oulasvirta, & Howes, 2015). The model uses reinforcement learning (RL) to compute the optimal, utility-maximizing policy.…”