Challenges in Ubiquitous Computing Patrick Malatack

Today’s Readings  Charting Past, Present, and Future Research in Ubiquitous Computing  by Gregory Abowd and Elizabeth Mynatt  The Origins of Ubiquitous Computing Research at PARC in the Late 1980s  by Mark Weiser, Rich Gold, John Seely Brown  Some Computer Science Issues in Ubiquitous Computing  by Mark Weiser  Charting Past, Present, and Future Research in Ubiquitous Computing  by Gregory Abowd and Elizabeth Mynatt  The Origins of Ubiquitous Computing Research at PARC in the Late 1980s  by Mark Weiser, Rich Gold, John Seely Brown  Some Computer Science Issues in Ubiquitous Computing  by Mark Weiser

Charting Past, Present, and Future Research in Ubiquitous Computing  Natural interfaces  Context-aware applications  Automated capture and access  Continuous interaction  Privacy  Security  Visibility  Control  Natural interfaces  Context-aware applications  Automated capture and access  Continuous interaction  Privacy  Security  Visibility  Control

Computing with Natural Interfaces  Ubicomp inspires off the desktop applications  Needs off the desktop means of interaction  Speech, gestures, writing  More accessible  Easier to use???  Ubicomp inspires off the desktop applications  Needs off the desktop means of interaction  Speech, gestures, writing  More accessible  Easier to use???

Problems with Natural Interfaces  First Class Data Types  Keyboard and mouse  Characters and x,y coords  Most natural interfaces try to convert to first class type  Pen to text, speech to text  Current Research is working with freeform input as a primitive.  Need to find other ways of structuring input  What other problems does creating new primitive types impose  Development issues  First Class Data Types  Keyboard and mouse  Characters and x,y coords  Most natural interfaces try to convert to first class type  Pen to text, speech to text  Current Research is working with freeform input as a primitive.  Need to find other ways of structuring input  What other problems does creating new primitive types impose  Development issues

Problems with Natural Interfaces  Error Prone Interaction  Permit new and numerous mistakes  People do not have perfect recognition  People don’t recognize increases in accuracy  Needs to be 5-10% to be recognized  Three areas  Error Reduction  Error Discovery  Reusable toolkit for error handling  Error Prone Interaction  Permit new and numerous mistakes  People do not have perfect recognition  People don’t recognize increases in accuracy  Needs to be 5-10% to be recognized  Three areas  Error Reduction  Error Discovery  Reusable toolkit for error handling

Questions about Natural Interfaces  Can we ever capture subtleties of natural human communication??  What different natural interfaces might exist??  Nintendo Wii  Trauma Center, Wii Sports  iPhone  How is accidental communication handled?  StarTrek  “Computer:play music”  Why do developers want to build apps for these interactions?  More errors, more edge cases  Can we ever capture subtleties of natural human communication??  What different natural interfaces might exist??  Nintendo Wii  Trauma Center, Wii Sports  iPhone  How is accidental communication handled?  StarTrek  “Computer:play music”  Why do developers want to build apps for these interactions?  More errors, more edge cases

Context Aware Computing  Current Systems  Generally using position and identification of objects  Still do not provide a complete context  Definition of context is limited  Research Areas  Context Toolkits  Toolkit for sensing environment  Explicit use of sensed information is up to program  What is Context?  How is Context represented?  Current Systems  Generally using position and identification of objects  Still do not provide a complete context  Definition of context is limited  Research Areas  Context Toolkits  Toolkit for sensing environment  Explicit use of sensed information is up to program  What is Context?  How is Context represented?

What is Context?  Who  Currently generally tailored to one user  How important are others in determining our behavior  How could this be captured?  What  Attempt to figure out what is currently happening  Sense environment, use calendar software etc.  What is the cost of the wrong guess?  TiVo Problem  Where  Location based information  GPS  Most explored of context information  When  Easily obtained information - Computer good a remembering time  Although determining when one event stops and another begins is not easy  Why  Even harder than the “what” question, biometric sensors might help  Who  Currently generally tailored to one user  How important are others in determining our behavior  How could this be captured?  What  Attempt to figure out what is currently happening  Sense environment, use calendar software etc.  What is the cost of the wrong guess?  TiVo Problem  Where  Location based information  GPS  Most explored of context information  When  Easily obtained information - Computer good a remembering time  Although determining when one event stops and another begins is not easy  Why  Even harder than the “what” question, biometric sensors might help

Other Problems with Context  How do computers represent context  No universal context schemes, toolkits must be developed with standard context representations  Context may contain a lot of information with only a few pieces relevant for determining context.  Context Sensing and Fusion  Sensors are not as flexible as human sensors  Right sensor for right job  Use multiple sensors in parallel to offset noise  Bit of a chicken-and-egg problem  How do you know when to switch sensor? Use context  How do computers represent context  No universal context schemes, toolkits must be developed with standard context representations  Context may contain a lot of information with only a few pieces relevant for determining context.  Context Sensing and Fusion  Sensors are not as flexible as human sensors  Right sensor for right job  Use multiple sensors in parallel to offset noise  Bit of a chicken-and-egg problem  How do you know when to switch sensor? Use context

Automated Capture and Access  Recording information and data as it occurs  Computers are inherently good at recording, people are not  People freed up to summarize and understand  Most work in academic/ classroom settings  Time stamping lectures, digital whiteboards  Challenges in capture and access  Sometime we don’t know we want to capture something until after its already happened  How could the computer know that?  If it captures everything then we need a system of sorting and filtering (access)  Gmail Problem  Access is a problem because capturing of raw data can be burdensome for sifting through…systems need to recognize important events facilitate access  Recording information and data as it occurs  Computers are inherently good at recording, people are not  People freed up to summarize and understand  Most work in academic/ classroom settings  Time stamping lectures, digital whiteboards  Challenges in capture and access  Sometime we don’t know we want to capture something until after its already happened  How could the computer know that?  If it captures everything then we need a system of sorting and filtering (access)  Gmail Problem  Access is a problem because capturing of raw data can be burdensome for sifting through…systems need to recognize important events facilitate access

Challenges Everyday Computing  No clear beginning and end  Both fundamental activities like communication and Long-term endeavors do not have predefined starts and ends  Traditional HCI deals with transition users from current state to a goal state  Not necessarily a goal state  Interruptions  People are constantly interrupted  Constantly on computer systems must recognize interruption and change state  Also computers must appropriately interrupt users  Concurrency  People multitask and rapidly switch task based on external unpredictable environment  Systems need to adapt to this opportunistic behavior and change accordingly  No clear beginning and end  Both fundamental activities like communication and Long-term endeavors do not have predefined starts and ends  Traditional HCI deals with transition users from current state to a goal state  Not necessarily a goal state  Interruptions  People are constantly interrupted  Constantly on computer systems must recognize interruption and change state  Also computers must appropriately interrupt users  Concurrency  People multitask and rapidly switch task based on external unpredictable environment  Systems need to adapt to this opportunistic behavior and change accordingly

Research in Everyday computing  Continuously present interface  No current model of continuously present interfaces, even people are not continuously present  Create an interface that doesn’t get annoying  Determine what information should require my attention and what should be display peripherally  Comfort and style  If constantly worn or carried must be stylish and comfortable  Continuously present interface  No current model of continuously present interfaces, even people are not continuously present  Create an interface that doesn’t get annoying  Determine what information should require my attention and what should be display peripherally  Comfort and style  If constantly worn or carried must be stylish and comfortable

Other Challenges for Ubicomp  Hard to evaluate Ubicomp Systems  Little publish on ubicomp evaluation  Systems often required to be fully connected leading to systems that are hard to build  Lack of development toolkits make system creation difficult  Systems often need to be integrated into peoples lives which using big clunky prototypes does not lead itself well too  Task/Goal centric approaches don’t work in ubicomp  Hard to evaluate Ubicomp Systems  Little publish on ubicomp evaluation  Systems often required to be fully connected leading to systems that are hard to build  Lack of development toolkits make system creation difficult  Systems often need to be integrated into peoples lives which using big clunky prototypes does not lead itself well too  Task/Goal centric approaches don’t work in ubicomp

Social issues in Ubicomp  Privacy  Potential for misuse  How should users be informed of computer monitoring (Don’t be invisible??)  Is it ok to give up this privacy in some settings, healthcare  Do we give the hackers too much power  Security  Constant monitoring take away personal responsibility  What happens if I lose my data  Privacy  Potential for misuse  How should users be informed of computer monitoring (Don’t be invisible??)  Is it ok to give up this privacy in some settings, healthcare  Do we give the hackers too much power  Security  Constant monitoring take away personal responsibility  What happens if I lose my data

Killer App in Ubicomp  Article talked about Ubicomp as killer existence rather than just killer app  That said killer apps like spreadsheets and word processing drove desktop proliferation  Does Ubicomp need a killer app to spur is adoption?  Where would such a killer app come from?  Business setting or home users  Article talked about Ubicomp as killer existence rather than just killer app  That said killer apps like spreadsheets and word processing drove desktop proliferation  Does Ubicomp need a killer app to spur is adoption?  Where would such a killer app come from?  Business setting or home users