Description
Model the activity a sensor observes and the task the system helps with, then match intervention desirability to sensing reliability to choose sensors, thresholds and fixes.
For systems that act on things people do for other purposes (lights that come on as a door opens, a sign that updates when a cafe opens), Dix separates the sensed task, the user's own activity as observed by sensors or logs, from the supported task the system aids. The supported task is annotated step by step with how wanted an intervention is (++ to --, with a special mark for safety issues). The sensed task is annotated with available or addable sensors and how likely each sensed state is to match reality. A state diagram grouped by what sensors can distinguish shows likely and unlikely transitions. The analysis aims to get critical cases right and others good enough, and finds states that are both wrong and likely, which are then made less likely or more noticeable.
Attributes
- Discipline
- Task Analysis & Cognitive Modelling
- Loss type
- Latent
- Stage
- Definition, Design
- Data
- Analytical / expert
- Effort
- Days
Steps
- Describe the supported task and the sensed task (scenario, HTA or CTT) and model the states of relevant objects.
- Annotate each supported-task step with desirability of intervention (++, +, -, --, safety) and classify requirements as absolute, strong or nice to have.
- Annotate the sensed task with existing and addable sensors and the likelihood that each sensed state matches the real state.
- Draw system states grouped by what sensors can distinguish, marking likely and unlikely transitions and where intervention is wanted.
- Match desirability against reliability: make critical cases correct, allow leeway elsewhere, and add timers or sensors where states cannot be told apart.
- List states that are both wrong and likely, and redesign to make them less likely or more noticeable.
- Review privacy on the sensed task and explanation on the supported task.
Sources
- The Handbook of Formal Methods in HCI (Weyers et al., 2017)
- Touch IT (Dix et al., 2022)
Also known as
- Sensed-task / supported-task analysis for low-intention interaction
- Sensed-task and supported-task modelling for low-intention interaction
- Incidental interaction analysis
- Two-task analysis
- Low-intention interaction design analysis
- Desirability versus sensing reliability analysis
Categories
- User Research
- Task Analysis & Cognitive Modelling