This paper presents a Monte Carlo method for processing uncertain input probabilistically.
(“This paper presents a set of techniques for tracking the state of interactive objects in the presence of uncertain inputs”)
Users’ input is taken as samples, weighted and dispatched to interactive objects whose state machines process this input, resulting in a set a sampled states and optional action requests (potentially to be executed). Then a mediation process handles the decision of rejecting/deferring/accepting certain action requests before finally execute them on the interface.
- In this paper, we focus on how to interpret uncertain input with respect o interactor state and on giving appropriate feedback to accurately reflect this uncertainty.
- Monte Carlo methods span a range of specific techniques but share the property that probability distributions are approximated by a set of samples over that distribution.
- Can we consider the ‘pipeline’ of interaction? In some sense, certainty is the ‘tip’ or the final phase of interaction (one interaction move ends after it) but uncertainty expands this ‘tip’ to a phase with fuzzy boundaries – even irrelevant input is considered to tolerate the ambiguity. Such ambiguity might be due to an intermediate phase of the process where the user tries to accomplish something. Norman’s stages might help explain some. While both certainty and uncertainty are explicit, how about the implicit – input that cannot be measured by (un)certainty?
- After all, how much have the uncertainty-aware techniques improve? Does it really make a difference?
- Very good observation of new/old technologies: ‘however, new interaction technologies violate the standard assumption that input is certain‘