At Home with Ubiquitous Computing: Seven Challenges W. Keith Edwards and Rebecca Grinter UbiComp 2001 Overview • “Smart homes” better people’s lives with increased communication, awareness, and functionality • However, there exist technical, social, and pragmatic challenges • Raise awareness of existing domestic technology literature • Increase the use of situated studies The smart home is coming • Technology is getting there: – Moore’s Law – Everything networked and wireless – Increased vendor focus on techs for the home – Proof-of-concepts exist: Aware Home @ Ga Tech • But there still exist some challenges! Seven challenges • • • • • The “accidentally” smart home Impromptu interoperability No systems administrator Designing for domestic use Social implications of aware home technologies • Reliability • Inference in the presence of ambiguity 1: The “accidentally” smart home • Current smart home environments are intentional (purpose-built) • More realistic view: technology will be brought piecemeal into the home (upgrade) – The “accidentally” smart home 1: The “accidentally” smart home • Even mundane examples, demonstrate big problems • How do users debug their home? • Is this simply a “design” problem? 1: The “accidentally” smart home • Solution is to help users to understand the tech – What devices can do, what they have done, and how we control? • When designing, think of these questions: – What kinds of affordances (action possibilities, e.g., recording, displaying) do we need to make the system intelligible? – How can I tell my device is interacting? – What are the boundaries of my smart home? – What are the potential configurations of my devices? – How can users be made aware of the entire houses’ affordances? – Where will the locus of interaction be in a system that isn’t in one place (but sum of many parts)? – How do I control these devices and the whole system? 2: Impromptu interoperability • Ability to interconnect with little advance planning • A priori agreement on syntax and semantics is needed • However, creating standards for all types of devices/services (a priori) is not feasible • New models of interconnectivity are required 3: No systems administrator • Can’t plausibly expect that homeowners will need to be system administrators • How about “appliance-centric” computing (single function oriented)? Still having interoperability problems? • Utility model: “thin-client” solution?? – Open services gateway initiative – Cloud computing • Why doesn’t plumber/electrician model work? 4: Designing for domestic use • Learning from the telephone/autos/cell-phones – Hard to foresee how people use a tech (intention vs. actual use) • Learning from domestic technology studies – Domestic technology use governed by rules of the house – Television use indicated who “controlled” an area of the house – Teenagers used individually owned technology to coordinate using a shared technology (e.g., “quiet” technologies to avoid disrupting other’s routines) • Designers need to pay attention to the subtle house routines + how occupants adapt new techs? 5: Social implications of aware home technologies • Social implications of domestic technologies • Are domestic technologies labor saving? – Introduction of technology into the home changes societal expectations – Has the introduction of technology increased or shifted the amount of work you do? • TV has changed “good parenting” to controlling what not if your child watches – In Europe, mobile phones teaches children about managing money and safely gives them increased independence 6: Reliability • Current domestic appliances are pretty reliable • Differences: domestic vs. desktop (& ubicomp?) – Development culture • Embedded vs. general-purpose? – Technological approaches • Phone (thin) vs. web surfing (thick) – Expectations of the market • Crashing washing machine vs. desktop? – Regulations • Highly regulated appliances (due to safety concerns) 7: Inference in the presence of ambiguity • Current machine inference is kind of bad (e.g. Microsoft Clippit) • How smart does a smart home have to be? • Is it better not to act, or to act and be wrong? • Modes of intelligence: – Infer state of world through interpretation of sensor data – Infer existence of states by aggregating other factors (e.g., people gathering at a meeting room --> meeting?) – Infer my intent from its view of the state of the world (e.g., meeting sharing notes with others) – Preemptively act on the assumptions of intent 7: Inference in the presence of ambiguity • Predictability is important (e.g., dropping temperature thermostat turns on the heating) • For a given condition, predictability depends on: – System’s expected behavior under the condition – System’s facilities for detecting/inferring the condition – Provision for user to override the system’s behavior • How can we redesign the Bluetooth speakers to be more predictable?