Key Moments
Stanford CS547 HCI Seminar | Autumn 2026 | Sensing Uncertainty in Sustainable HCI
Stanford OnlineKey Moments
Virtual humans are now nearly indistinguishable from real ones — but each one costs $1M to capture.
Key Insights
Sauna 4-7x/week reduces cardiovascular death risk by 50% vs once/week (BMC Medicine, n=1,688).
Google's Med-Gemini outperforms doctors at diagnosis, raising the question of whether AI should be deployed even when we can't explain its reasoning.
The average person uses 5-10 devices, but embedded IoT could mean 50-100 devices per person.
The UBench system, developed with wood artisans, helps users identify and manage material flaws directly on the workpiece.
Studies show static visualizations lead to more accurate interpretations of air quality data than interactive or animated ones.
Creative reuse of existing objects for IoT requires addressing material impurities and providing intuitive user interfaces for handling uncertainty.
The low-carbon innovation paradox fuels a need to embrace uncertainty
The push for sustainability in computing faces a paradox: technologies designed to reduce environmental impact can inadvertently exacerbate the problem. For example, the Internet of Things (IoT), while promising to optimize energy consumption through data collection, dramatically increases the number of devices per person. We've moved from one device per person to 5-10 with mobile and wearables, and embedded IoT could lead to 50-100 devices per person. This proliferation, from manufacturing to data processing, contributes to carbon emissions. Instead of abandoning these technologies, the speaker argues for a new approach that is more comfortable with uncertainty, moving beyond traditional analyses of short-term versus long-term benefits or 'rebound effects'. This new perspective focuses on the uncertainty inherent in the data itself and how user interfaces shape our understanding and use of technology. The core question becomes: are we really reporting on events as they occurred, or do we need to consider this data as potentially misleading and understand the range of possibilities?
Creative reuse: giving old objects a new life in the IoT
The speaker's research into 'creative reuse' explores how to integrate existing objects into the IoT ecosystem, contrasting with 'downcycling' where objects lose value. This approach values the history and social connections embedded in physical objects. A study with 10 families revealed that homeowners prefer not to replace reliable analog devices with potentially fragile digital counterparts. They found that the physical form and history of objects play nuanced roles in household dynamics, fostering creativity and supporting evolving interests. For instance, a well-used physical cookbook, with its stains and notes, offers a richer experience than a digital recipe. The challenge lies in designing IoT that acknowledges the 'messiness' of home life, rather than trying to automate it away. This involves supporting everyday organization, task handover, and people's mental models of their homes, which are often tied to the physical arrangement and use of objects. This perspective values the material history and social networks that objects are part of, allowing them to participate in an ecosystem without necessarily becoming 'smart' in the conventional sense.
Bridging the gap for artisans in the circular supply chain
To enable 'creative reuse' within circular supply chains, the speaker's lab investigates methods borrowed from 'garbology'—the study of material culture and waste. A significant design challenge is equipping people with the skills and technologies to work with used materials, which often have unpredictable properties. Unlike new materials with uniform characteristics, salvaged items like reclaimed wood may contain nails, toxic coatings, or structural inconsistencies. To address this, they collaborated with four experienced wood artisans. The process involved observing their workflows, using a 'generative toolkit' to extract implicit knowledge, and conducting semi-structured interviews about how they incorporate material imperfections into their designs. Key findings from this work led to the design logic of a system that supports awareness of risks, manages decisions made about these risks, and allows for direct, hands-on work with materials rather than through CAD models. The goal is to embed expert strategies into tools, enabling learners to simulate expert practices and become more adept at handling material defects.
UBench: hands-on design with embodied material knowledge
Responding to the artisans' insights, the UBench system was developed to support hands-on creative reuse. It integrates sensing-based tools that allow for defect detection, interactive learning about material risks, and direct design on the workpiece, ensuring feedback is material-aware. The wooden piece itself acts as the interactive surface, grounding evaluation, interpretation, and design. An overhead camera module assesses the wood, projecting defect information back onto the material. Sensing pens can detect hidden flaws like metal or moisture, with their positions tracked in 3D. A metal-detecting pen uses an inductive sensor, while a moisture sensor detects risks from humidity. This information is then documented and displayed on the wood, spatially linking defects to the workpiece. For on-site design, a traceable pen cap allows users to draw directly on the wood, with computer vision enabling interactive dialogue between the system and the user's design intent. When a user's design intersects with a defect, the system offers contextual suggestions based on the defect type and chosen strategy, projecting these onto the wood to guide the artisan.
Visualizing uncertainty in air quality data
The research shifts to sensing uncertainty in data, specifically focusing on air quality. Sensor data is inherently uncertain due to issues like sensor drift caused by environmental changes or wear. This uncertainty can mislead decision-makers. To explore how interaction design can support this, nine data visualizations were created, varying in interactivity (static, animated, interactive) and context (spatial, temporal, sensor-only). These visualizations represented PM 2.5 data converted to US EPA Air Quality Index (AQI) levels. Findings revealed that static visualizations supported more accurate interpretations of AQI, aligning with expectations. However, interactive and animated visualizations, while better for comparing sensor reliability, surprisingly led to more inaccurate AQI interpretations. Spatial context, contrary to expectations, also resulted in less accurate answers than temporal or sensor-only contexts. The speaker posits that the design of visualizations, particularly the use of gradients versus sharp color breaks, might influence how users perceive and interpret uncertainty, potentially leading to a trade-off between accurate AQI interpretation and understanding data uncertainty.
Designing for uncertainty: trade-offs and future directions
The study on air quality visualization highlighted a crucial trade-off: static visualizations facilitate accurate AQI interpretation but are less effective for understanding data uncertainty and reliability comparisons. Conversely, interactive visualizations excel at comparing sensor reliability but can lead to misinterpretations of AQI. This suggests that precisely because we want users to understand uncertainty (e.g., during COVID-19), we might need designs that make this uncertainty explicit, even if it leads to less precise immediate interpretations. Future directions include exploring emerging technologies like batteryless IoT devices, which, despite reducing toxic waste, rely on inherently inaccurate sensors. This necessitates developing user rules for interpreting noisy sensor data and designing tangible interactions that make this data accessible. Another challenge is aligning users' mental models with data feedback, particularly when resource recovery from e-waste is considered. The speaker emphasizes that instead of expecting everyone to become sustainability experts, new forms of interaction design are needed to help people navigate uncertainty, fostering a collective response and creating opportunities for community-level problem-solving.
Mentioned in This Episode
●Products
●Software & Apps
●Organizations
●Concepts
●People Referenced
Common Questions
The video discusses the 'carbon innovation paradox,' where technologies intended to help with sustainability might actually worsen the problem due to increased device proliferation, manufacturing, and energy consumption for data processing.
Topics
Mentioned in this video
An ethnographic and archaeological field of study that informs new approaches to understanding people's relationship with material culture and waste.
A spatial interpolation algorithm used to estimate pixel color at a location where no direct sensor reading was available, discussed as potentially unsuitable for certain air quality data scenarios.
Discussed as a technology that promises energy efficiency but contributes to increased device proliferation and carbon emissions.
Fine particulate matter used in the study to convert sensor readings into an Air Quality Index based on EPA guidelines.
More from Stanford Online
View all 160 summaries
79 minStanford AA274A Principles of Robotic Autonomy | Autumn 2019 | EKF Localization
78 minStanford AA274A Principles of Robotic Autonomy | Autumn 2019 | Parametric Filtering
57 minWebinar: Regulating at the Speed of AI with Francis Fukuyama and Nate Persily
78 minStanford AA274A Principles of Robotic Autonomy | Autumn 2019 | Localization & Filters
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free