Assistive interfaces, such as recommendation engines, adaptive systems, and intelligent assistants, span diverse methods and disciplines but lack a shared conceptual foundation. This paper models assistance as sequential decision-making under uncertainty between two agents: the user and the assistant. The formalism allows casting assistance as an optimization problem and offers a rich but principled vocabulary to understand the dynamics of assistance. Drawing on Partially Observable Stochastic Games (POSGs) and related models, we: (1) motivate multi-agent over single-agent formulations; (2) adapt POSGs to HCI and clarify their tractability through reductions; (3) propose a two-agent sequential model that unambiguously defines concepts such as adaptation, augmentation, and delegation; (4) illustrate applicability through domain problems and examples; and (5) offer a supporting implementation via a library. These results warrant more attention on decision-theory as a principled yet actionable approach to assistive interfaces.

This article tackles two challenges of the empirical evaluation of interaction techniques that rely on user memory, such as hotkeys, here coined Recall-based interaction techniques (RBITs): (1) the lack of guidance to design the associated study protocols, and (2) the difficulty of comparing evaluations performed with different protocols. To address these challenges, we propose a model-based evaluation of RBITs. This approach relies on a computational model of human memory to (1) predict the informativeness of a particular protocol through the variance of the estimated parameters (Fisher Information) (2) compare RBITs recall performance based on the inferred parameters rather than behavioral statistics, which has the advantage of being independent of the study protocol. We also release a Python library implementing our approach to aid researchers in producing more robust and meaningful comparisons of RBITs.

Fitts’ law is a behavioral model, used to design protocols and analyze data from pointing experiments. These are usually conducted in HCI to evaluate input performance. We recently proposed an alternative method to characterize input performance, called the method of PVPs in 1D, based on 1) a dual-minimization protocol, and 2) an analysis of the variability of entire trajectories. We extend the method in 2D; our contributions include new metrics, a new protocol, and a Python library. We also present the results of a controlled experiment where the new method is validated using three devices (mouse, touchpad, controller): effect sizes in the 2D case replicate those previously found. We also propose a comparison between Fitts’ law and our novel evaluation: the method of PVPs provides more information than Fitts’ law, and can predict its parameters. We discuss how this new method may relieve open problems of Fitts’ law.

Hick's law is a key quantitative law in Psychology that relates reaction time to the logarithm of the number of stimulus-response alternatives in a task. Its application to HCI is controversial: Some believe that the law does not apply to HCI tasks, others regard it as the cornerstone of interface design. The law, however, is often misunderstood. We review the choice-reaction time literature and argue that: (1) Hick's law speaks against, not for, the popular principle that 'less is better'; (2) logarithmic growth of observed temporal data is not necessarily interpretable in terms of Hick's law; (3) the stimulus-response paradigm is rarely relevant to HCI tasks, where choice-reaction time can often be assumed to be constant; and (4) for user interface design, a detailed examination of the effects on choice-reaction time of psychological processes such as visual search and decision making is more fruitful than a mere reference to Hick's law.

Knowledge management and sharing involves a variety of specialized but isolated software tools, tied together by the files that these tools use and produce. We interviewed 23 scientists and found that they all had difficulties using the file system to keep track of, re-find and maintain consistency among related but distributed information. We introduce FileWeaver, a system that automatically detects dependencies among files without explicit user action, tracks their history, and lets users interact directly with the graphs representing these dependencies and version history. Changes to a file can trigger recipes, either automatically or under user control, to keep the file consistent with its dependants. Users can merge variants of a file, e.g. different output formats, into a polymorphic file, or morph, and automate the management of these variants. By making dependencies among files explicit and visible, FileWeaver facilitates the automation of workflows by scientists and other users who rely on the file system to manage their data.

The design of Fitts' historical reciprocal tapping experiment gravely confounds index of difficulty ID with target distance D: Summary statistics for the candidate Fitts model and a competing model may appear identical, and the validity of Fitts' model for some tasks can be legitimately questioned. We show that the contamination of ID by either target distance D or width W is due to the common practices of pooling and averaging data belonging to different distance-width (D,W) pairs for the same ID, and taking a geometric progression for values of D and W. We analyze a case study of the validation of Fitts' law in eye-gaze movements, where an unfortunate experimental design has misled researchers into believing that eye-gaze movements are not ballistic. We then provide simple guidelines to prevent confounds: Practitioners should carefully design the experimental conditions of (D,W), fully distinguish data acquired for different conditions, and put less emphasis on r² scores. We also recommend investigating the use of stochastic sampling for D and W.

Human movements are inherently variable and involve some feedback mechanism. A study of the positional variance in a tapping task reveals that the variance profiles are unimodal in time. In the variance-decreasing phase, the aiming task can be modeled by a Shannon-like communication scheme where information is transmitted from a "source" - determined by the distance to the target at maximum variance - to a "destination" - the movement endpoint - over a "channel" with feedback perturbed by Gaussian noise. Thanks to the feedback link, we show that the variance decreases exponentially at a rate given by the channel capacity. This is confirmed on real data. The proposed information-theoretic model has promise to improve our understanding of human aimed movements.

The rationale for Fitts’ law is that pointing tasks have the information-theoretic analogy of sending a signal over a noisy channel, thereby matching Shannon’s capacity formula. Yet, the currently received analysis is incomplete and unsatisfactory: There is no explicit communication model for pointing; there is a confusion between central concepts of capacity (a mathematical limit), throughput (an average performance measure), and bandwidth (a physical quantity); and there is also a confusion between source and channel coding so that Shannon’s Theorem 17 can be misinterpreted. We develop an information-theoretic model for pointing tasks where the index of difficulty (ID) is the expression of both a source entropy and a zero-error channel capacity. Then, we extend the model to include misses at rate ε and prove that ID should be adjusted to (1−ε)ID. Finally, we reflect on Shannon’s channel coding theorem and argue that only minimum movement times, not performance averages, should be considered.

In usual Fitts' law experiments the outcome of a pointing act can be either measured as an error, i.e., a distance from endpoint to target center, or categorized in an all-or-none way as a hit versus a miss. Information theory offers a useful distinction between transmission errors (the received symbol is wrong) and erasures (the received symbol is empty). Although Fitts' law research has been very much inspired by the information theoretic rationale, the error/erasure distinction has escaped attention so far: Target misses have always been treated as normally-distributed errors, through the effective index of difficulty IDe. The paper introduces a new index of difficulty based on the simple observation that a target miss conveys zero bit of information, i.e., it is an erasure. Not only is the new index more consistent with the fundamentals of information theory, it is much simpler to derive than the ISO-recommended IDe.