ContextualIntelligence Bo Begole, Ph.D.Samsung Research America, UX Centerbo@begole.netKorean HCI Society Conference31 January 2013
Businesses strategies for technologiesthat bridge physical and digital worlds• Capabilities, Techniques,Limitations• How to exploit the trends• Key value propositions• Case Studies in• Contextual Intelligence• Hyper-Personal informationfiltering• Predictive personal marketing• Supply chain disruption• Unanticipated deviceinteroperation• Consumer decision supportFT Press, ISBN-13: 978-0-13-706443-4
20 Years ago, Ubiquitous Computingwas a dream …3Olivetti ActiveBadgePARC Tab PARC Pad PARC “Deathstar” IRNetwork Hub“The most profound technologies arethose that disappear.They weave themselves into thefabric of everyday life until they areindistinguishable from it.”-Mark Weiser, Xerox PARChttp://www.youtube.com/watch?v=b1w9_cob_zw
Too Many Devices!Too Much Information!Too Many Friends!Too Many Networks!Too Much To Do!
Context-Aware services reducetechnology overloadSystems detec users’ locationsto help them– Find colleagues– Find resources– Seamlessly connect tonearby devices/services– Route phone calls– Tag information creationby context for recall– Diary of daily activities (1994) Bill Schillit, Norm Adams, Roy WantContext-Aware Computing ApplicationsWorkshop on Mobile Computing Systems and Applications,6
Context-aware is now table-stakes. Newfrontier: Behavior, Activity and Goal-AwareContext-AwareSensor-triggered services withno intelligence:Current/future location; timeKey: distance in time/space9IncreasingIntelligence1990-2004Predictive analytics frombehavior traces:rank info by frequency/recency ofpeople, apps, docs, locations; …Key: distance from “normal”Behavior &Activity-Aware2005-2015Semantic understandingof workflows and objectives:writing a proposal; filling sales pipeline;shopping for gift; fostering relationships;…Key: distance from “objective”Goal-Aware2015+
Sternberg’s Tri-archic Theory ofIntelligence• Analytic Intelligence– Logic, sequences, calculations, …• Synthetic Intelligence– Creativity, intuition, …• Contextual Intelligence– Practical “street smarts”– Understanding relationships and utility– Achieve goals with available resources10
Contextually Intelligence Assistant:Pertinent and Actionable information11http://youtu.be/FPL7uaKYNiU
Nice vision, but it’s been 20 years…• Some classic problems are still unsolved…– Privacy– Missing information• … and bigger problems have arisen– Latencies• Delays to build model• Delays in detecting pattern change• Delays to detect situation– Accuracy: typically 85% at best– Actionable? Tell me something I don’t know12
Privacy Concerns• Security and access control isonly half the story• Impression Management– Confusing interfaces,contradicting policies– What is this thing telling otherpeople about me?• Economic Risk– Unwanted sales pressure– Higher insurance rates13
Current digital assistants have to askfor every detail14“The address wasthere in my contactcard, but Siri kepttelling me, ‘I don’tknow where yourhome is.’”Why do I have to tell it everything?Why can’t it figure things out and confirm?
Latencies• Pattern models take time to build– Frequent patterns build quickly– Weekly, monthly, annual patterns build slowly• Changes take time to detect• Recognition is often slower than humanintelligence15GPSNFCEMAILCalendar
Accuracy – best are wrong 15-20%16• Fine for advertising or product recommendations• What is good enough for content retrieval?
OptimalTimingActionContextReminders inactionablecontextPertinent Information& Actionable Contextn Communication: tell mebefore someone becomesunavailablen Health & Fitness:suggestions of healthyoptions not already inpracticen Technician: take preventiveparts before repair visitn Advertisement: Deliver adswhen person is actually “inthe market”• Health plan• Task list• Reminders• Coupons
Technology alone cannot solve these• Privacy – Requires understanding real user fears• Missing information – Human conversational repair• Latencies – Give users control of their pattern models• Accuracies – Adjust UX depending on application need• Actionability – Don’t be a nag. What do they already know?What will achieve their goal? How is their goal defined?HCI Grand Challenge:HCI research methods can create the knowledgeto emulate human contextual intelligence18