Description
A concept-stage paper canvas that links the UX challenges an ML component creates to the user, scenario and ML-system factors behind them, to surface design opportunities.
The ML-Process Canvas helps designers think about machine learning during conceptual design without abandoning their usual activities. It first prompts the designer to name the UX challenges that ML introduces, such as uncertain or wrong outputs, then to describe the user, the scenario and the elements of the ML system (data, model, feedback) that could affect each challenge. Findings are organised on the canvas linked to those elements, and the designer reads across for design opportunities; in the published example, system responses were kept low-key because the model could not guarantee correct outputs. The later ML Lifecycle Canvas folds the ML development process into conceptual design, treating ML, users and situation as one whole and giving designers and data scientists a shared artefact.
Attributes
- Discipline
- Design & Product Practice
- Loss type
- Latent
- Stage
- Definition, Design
- Data
- Analytical / expert
- Effort
- Hours
Steps
- Use the canvas during conceptual design alongside normal research synthesis.
- Identify the specific UX challenges the ML component creates (for example error, uncertainty, cold start, opacity).
- Describe the user, the scenario and the ML system elements (data, model, learning loop) that could affect each challenge.
- Organise findings on the canvas linked to those elements, ideally with a data scientist present.
- Read across the canvas for design opportunities and record candidate responses, such as softening outputs where confidence is low.
Sources
- User Experience Design Using Machine Learning (Abbas et al., 2022)
Also known as
- ML-Process Canvas
- ML Lifecycle Canvas
- ML-Process Canvas and ML Lifecycle Canvas
Categories
- Design
- Design & Product Practice