MusicFX: MusicFX: An Arbiter of Group An Arbiter of Group Preferences for Computer Preferences for Computer Supported Collaborative Supported Collaborative Workouts Workouts Joe McCarthy Joe McCarthy Ted Ted Anagnost Anagnost Andersen Consulting Andersen Consulting Center for Strategic Center for Strategic Technology Research Technology Research
31
Embed
MusicFX: An Arbiter of Group Preferences for Computer Supported Collaborative Workouts (CSCW98)
MusicFX is an example of an active environment that uses a group preference arbitration system to allow the members of a fitness center to influence, but not directly control, the selection of music in that environment. The system contains a database of members' musical preferences, a badge system for determining who is working out, and a weighted random selection algorithm for selecting music to best suit the group inhabitants at any given time. MusicFX was deployed in the fitness center at Accenture Technology Park in Northbrook, IL (USA) from November 1997 through January, 2002. These slides are from the CSCW 98 presentation on the system. More info, including the CSCW 98 paper, can be found at http://interrelativity.com/joe/projects/MusicFX.html
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
MusicFX:MusicFX:An Arbiter of Group Preferences for An Arbiter of Group Preferences for
• ““UbiComp in a box”UbiComp in a box”– Interconnected, cooperating devicesInterconnected, cooperating devices– Concentrated in a small area (e.g., one room)Concentrated in a small area (e.g., one room)
Observation 1Observation 1• Most UbiComp applications focus onMost UbiComp applications focus on
single individualssingle individuals in in multiple spacesmultiple spaces– Active Badge: open doors, teleportingActive Badge: open doors, teleporting– ParcTab: information access, emailParcTab: information access, email
• What about What about multiple inhabitantsmultiple inhabitants in a in asingle, shared spacesingle, shared space??
Observation 2Observation 2
Music in the Fitness Center (FX)Music in the Fitness Center (FX)• Popular in the Complaint DepartmentPopular in the Complaint Department
– 25% of “feedback” focused on music25% of “feedback” focused on music
• RSI: Repetitive Song Injury RSI: Repetitive Song Injury – 3 stations played, 91 available (DMX)3 stations played, 91 available (DMX)
• Squeaky WheelsSqueaky Wheels– Vocal minority prevails over silent majorityVocal minority prevails over silent majority
• HangoversHangovers– This morning’s music = last night’s musicThis morning’s music = last night’s music
Four Issues for anyFour Issues for anyIntelligent EnvironmentIntelligent Environment• Who’s here?Who’s here?
• What are they doing?What are they doing?
• What are their preferences?What are their preferences?
• What can I do to help?What can I do to help?
Four Issues for MusicFXFour Issues for MusicFX• Who’s here?Who’s here?
– Members who login [badge reader]Members who login [badge reader]
• What are they doing?What are they doing?– Working out while listening to musicWorking out while listening to music
• What are their preferences?What are their preferences?– Diverse (to say the least)Diverse (to say the least)
• What can I do to help?What can I do to help?– Play “good” musicPlay “good” music
The MusicFX SystemThe MusicFX System
• Database of musical preferencesDatabase of musical preferences
• Group Preference Arbitration algorithmGroup Preference Arbitration algorithm– Group Preference CalculationGroup Preference Calculation– Candidate IdentificationCandidate Identification– Weighted Random Selection operatorWeighted Random Selection operator
Music Preference DatabaseMusic Preference Database• 275 fitness center members275 fitness center members• 91 musical genres (DMX stations)91 musical genres (DMX stations)• 5-point rating scale5-point rating scale
+2+2 = I = I lovelove this music this music +1+1 = I = I likelike this music this music 00 = I = I don’t minddon’t mind this music this music -1-1 = I = I dislikedislike this music this music -2-2 = I = I hatehate this music this music
Group Preference ArbitrationGroup Preference Arbitration
• Group Preference CalculationGroup Preference Calculation
• Weighted Random Selection Weighted Random Selection
Group Preference CalculationGroup Preference Calculation
N
jjii IPGP
1
2, 2
WhereWhere•GPGPii = = GGroup roup PPreference for genre reference for genre ii•IPIPi,ji,j = = IIndividual ndividual PPreference of person reference of person jj for genre for genre ii•NN = = NNumber of inhabitantsumber of inhabitants
Candidate IdentificationCandidate Identification
• Sort genre list by Sort genre list by GPGPii
• Remove any undesireable genre Remove any undesireable genre – Individual Preference FilterIndividual Preference Filter
• Candidates are the first M genreCandidates are the first M genre– Group Preference FilterGroup Preference Filter
Weighted Random SelectionWeighted Random Selection
• Calculate weights for candidatesCalculate weights for candidates
• Probabilistically select genre according to Probabilistically select genre according to WWii
M
iiGPTW
1 TW
GPW i
i
An exampleAn examplei Genre Person A B C D E GP i Pr i
1 Alternative Rock 2 2 0 2 2 68 0.48
2 Hottest Hits 1 1 2 0 -2 38 0.27
3 New Music 1 1 1 0 0 35 0.25
4 Hot Country 2 0 0 0 -2 28 0.00
5 Dance 2 -1 1 -1 -1 28 0.00
6 World Beat 0 1 -1 1 -2 23 0.00
7 Traditional Country 1 0 0 -2 -2 17 0.00
8 50's Oldies 0 0 0 -1 -1 14 0.00
9 Heavy Metal -1 -1 -1 -1 -2 4 0.00
10 Polka -1 -1 -2 -2 -2 2 0.00
Environmental EventsEnvironmental Events• Member entranceMember entrance
– Login (badge reader)Login (badge reader)
• Member exitMember exit– Timeout (90 minutes)Timeout (90 minutes)
– Individual / Group Preference Filter, Maximum Play TimeIndividual / Group Preference Filter, Maximum Play Time
• Maximum Play Time ElapsedMaximum Play Time Elapsed
The Success of MusicFX The Success of MusicFX • Daily operation since November 1997Daily operation since November 1997• Poll results (after 6 weeks)Poll results (after 6 weeks)
: increased variety, : increased variety, having some influence having some influence
: abrupt changes, : abrupt changes, occasional “bad” music occasional “bad” music
Better71%
Same8%
Worse7%
N/A14%
Evaluating Evaluating Group Preference ArbitrationGroup Preference Arbitration
• Calculate the “goodness” of MusicFXCalculate the “goodness” of MusicFX
• Estimate the “goodness” of old schemeEstimate the “goodness” of old scheme
• Compare the old with the newCompare the old with the new
The “Goodness” of MusicFXThe “Goodness” of MusicFX
Individual Satisfaction rating (Individual Satisfaction rating (ISIS))– TimeTimei,ji,j = time person = time person jj spent listening to genre spent listening to genre ii
– IPIPi,ji,j = person = person jj’s Individual Preference for genre ’s Individual Preference for genre ii
1
,
,, *i
n ji
jijij Time
IPTimeIS
The “Goodness” of MusicFXThe “Goodness” of MusicFX
Overall Satisfaction rating (Overall Satisfaction rating (OSOS))– For all For all NN members members
N
ISOS
j
Nj
1
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
IndividualIndividualSatisfactionSatisfactionfor all 275for all 275FX MembersFX Members
The “Goodness” of the Old DaysThe “Goodness” of the Old Days• Three genres (Three genres (n=3n=3))
– Hottest Hits, Power Hits, DanceHottest Hits, Power Hits, Dance– Assume each person listened to each genre 1/3 Assume each person listened to each genre 1/3
of the total time spent working outof the total time spent working out
1
3
,
3
*ijij
j
IPTotalTimeIS
Comparing the Old with the NewComparing the Old with the New