Stanford OnlineStanford CS547 HCI Seminar | Autumn 2026 | Sensing Uncertainty in Sustainable HCI
Stanford Online
Sustainability and uncertainty 0:10
Sustainability is not new, but the scale of the problem is. E-waste keeps growing, greenhouse gas emissions have not been curbed enough, and data centers and AI are driving new demand for energy and water. That creates a paradox. The tools meant to help with sustainability can also add to the burden through manufacture, networking, storage, and costly computation. The talk does not focus on long-term rebound effects. It focuses on sensing uncertainty: whether the data from a sensor really reflects what happened, or whether you need to treat it as a rough range. The speaker’s group studies this across the computing stack, from upcycled objects to user interfaces, so people can work with uncertainty instead of pretending it is not there.
Wood defects and risk 16:30
Four experienced secondhand woodworkers were studied in their workshops through contextual inquiry, recall of recent projects, and a toolkit for drawing out tacit knowledge about inspection and judgment. The work led to a simple design stance. You need to make risk visible early. You need to help people manage a chosen strategy. You need to let them work on the wood itself, not only in CAD. And you can build expert habits into the tools so learners can notice defects, judge them, and decide whether to feature, repair, or discard them while sketching directly on the material.
Uncertainty and accuracy 34:01
The speaker explains that the map work may be less accurate because of the gradient color scheme. A hard color break can be easier to read, but the team used gradients to show in-between states and uncertainty. Accuracy here means placing a reading into the EPA air-quality bins, not naming an exact value. People had to choose the right semantic band, even when the data itself was uncertain.
Design trade-offs 38:00
The results point to a trade-off. Static visualizations helped people make cleaner AQI judgments, while interactive ones helped them compare uncertainty and data reliability across sensors. The speaker says this may be why the interactive version looked worse for certainty and better for uncertainty. The wider point is that design can push users toward the expert view of uncertainty, but that view may not match how they naturally judge the data.
Upcycling and reuse 57:00
The discussion turns back to upcycling. Digital devices are harder to rethink than wood, so the speaker points to reclaimed parts from e-waste, modular layouts, and easy detachment without glue or solder. Batteries and other common failures may be replaced while other parts stay useful. The bigger question is what a second life should look like, and whether materials should move into new supply chains.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
