Improving User Experience By Employing Collective Intelligence April 16, 2008 Jason Zietz
Slide 2Meet the Presenter Education M.S. Software engineering and Application, Virginia Tech Thesis: Activity-based Knowledge Management Tool Design for Educators Work Experience Companies huge and little Currently User Experience expert
Slide 3Presentation Overview Background Examples of Collective Intelligence Implementing Collective Intelligence Applications in Current L3D Research
Slide 4What is Collective Intelligence? Aggregate insight is a type of knowledge that rises up out of the coordinated effort and rivalry of numerous people. (Wikipedia) Necessary Ingredients from Participants: Appropriate outlook Willingness to share Openness to the estimation of circulated knowledge for the benefit of all
Slide 5Why Do We Care About Collective Intelligence on the Web? Flag versus Clamor in the Long Tail
Slide 6Why Do We Care About Collective Intelligence Now?
Slide 7What Is User Experience?
Slide 8Why Do We Care About UX?
Slide 9Why Do We Care About UX?
Slide 10Why Is UX Important to Collective Intelligence (and the other way around)? Utility = Value/Effort " Reservoir of Goodwill " (Krug)
Slide 11Presentation Overview Background Examples of Collective Intelligence Implementing Collective Intelligence Applications in Current L3D Research
Slide 12Explicit versus Verifiable Activities Implicit Insight accomplished innately with no additional work from the client Explicit Insight requires particular movement from client
Slide 13Common Computer-based Collective Intelligence Applications Social Networks Discussion Forums Mailing Lists Rating Systems Tags
Slide 14Google Search A goliath proposal framework Condor (Gloor) Google Trends Asks: "What are individuals looking for?" Takes Google Search above and beyond
Slide 15Amazon System Activity Home page suggestions "Individuals who purchased this likewise acquired… " "Purchase this with this and get an extra 5% off" User Activity Item Viewing Purchasing "I Own It" Control (Yes/No) Rating System (1-5 Scale) Was this survey accommodating? (Yes/No) Tags
Slide 16Netflix Recommendations and the Netflix Prize $1,000,000 to contestant scoring 10% superior to Netflix's Cinematch proposal framework Began as a crowdsourcing try however turned into a wellspring of aggregate knowledge 12/2006 – Third place participant posted finish calculation online Netflix has joined thoughts from current pioneer into Cinematch Just a Guy in a Garage
Slide 17flickr
Slide 18flickr
Slide 19Other Examples Open Source Software del.icio.us – Social bookmarking by means of labeling Wikipedia – When crowdsourcing gets to be aggregate insight Digg Visualizations – Was UX overlooked?
Slide 20Presentation Overview Background Examples of Collective Intelligence Implementing Collective Intelligence Applications in Current L3D Research
Slide 21User Experience Tasks Requirements Gathering Task Flows/Wireframing/Prototyping Testing Evaluation
Slide 22Programming Collective Intelligence Using Tags Identification Searching Tag Clouds Not Using Tags UX Consideration
Slide 23Programming Collective Intelligence Making Recommendations Similarity Coefficients Euclidean Distance Pearson Correlation Tanimoto Similarity Score Others ( Jaccard , Manhattan , and so on) Cognitive Biases
Slide 24Euclidean Distance Used in appraisals frameworks Straight-line separate between two focuses Can be utilized to quantify distinction in evaluations by two individuals To get a comparability score between two individuals, figure which yields a number somewhere around 0 and 1, where 1 implies that the two individuals appraised the greater part of the things indistinguishably
Slide 25Pearson Similarity Coefficient Measure of how well two arrangements of information fit on a straight line Correlation of 1 means appraisals were indistinguishable
Slide 26Tanimoto Similarity Score Where N A : Total things in A N B : Total things in B N C : Total things in both An and B Tanimoto Similarity Score is the proportion of the crossing point set to the union set
Slide 27Cognitive Biases Psychological Effects That Can Skew Data Example: Anchoring in Netflix evaluations
Slide 28Clustering Prepare information utilizing normal arrangement of numerical ascribes used to look at things Choose grouping technique Hierarchical Clustering K-Means Clustering
Slide 29Hierarchical Clustering
Slide 30K-Means Clustering
Slide 31Clustering Blogs with Hierarchical Clustering
Slide 32Visualizing Clusters - Dendograms
Slide 33Clustering Blogs with Hierarchical Clustering
Slide 34Clustering Words inside Blogs with Hierarchical Clustering
Slide 35Presentation Overview Background Examples of Collective Intelligence Implementing Collective Intelligence Applications in Current L3D Research
Slide 36Applications to Meta-Design Meta-Design Explores Personally Meaningful Activities Output of aggregate knowledge applications must be important to members Meta-Design Requires Active Contributors Collective insight applications take into account an extensive variety of action, from certain to exceptionally unequivocal commitments Meta-Design Raises Research Problems, Including Collaboration and Motivation Collective insight applications can empower understood joint effort Collective knowledge applications can yield comes about generally not seen by members, along these lines expanding utility and decidedly impacting inspiration
Slide 37Applications in 'Transformative Models of Learning… " Why endeavor to enhance UX through Collective Intelligence in this examination? As the span of a VO scales upwards, the capacity to effectively recognize associations among individuals and find significant data diminishes Aiming to Create a VO of Active Contributors Utility = Value/Effort
Slide 38Applications in 'Transformative Models of Learning… " Link Members of VO Activity: Members of VO label themselves Tags – Skills they have, abilities they need (yet have use for), research premiums Use Tanimoto score to match individuals with comparable research premiums Use Tanimoto score to match individuals who do not have an aptitude with individuals who have that expertise
Slide 39Applications in 'Transformative Models of Learning… " Discover Relevant Areas of Study Activity: Rate coursework taken System stores past coursework of all members Students can rate this coursework as indicated by the amount they enjoyed the subject System utilizes appraisals to propose different zones of study which might enthusiasm to understudy
Slide 40Applications in 'Transformative Models of Learning… " Explore Relevant Content Activity: Cluster content inside VO Allow individuals from VO to investigate important substance in groups utilizing perceptions, for example, dendograms
Slide 41Applications in "… Using and Evolving Software Products" Increase Utility of SAP Message Boards Cluster related messages and permit clients to investigate the messages through an intelligent dendogram Make proposals of strings clients might be keen on perusing
Slide 42Suggested Readings Blog of Collective Intelligence (Pór) Programming Collective Intelligence (Segaran) Peter Morville on User Experience Design Elements of User Experience (Garrett) The Machine is Us/ing Us
SPONSORS
SPONSORS
SPONSORS
Changing the Classroom Experience: Face to Face Computer Supported Collaborative Learning Miguel Nussbaum mning 
Responsiveness in Father-Child Relationships: The Experience of Fathers 
Scattering and Use of Time Use Data The New Zealand Experience 
Interpersonal Knowledge Sharing: 
Spanning the CRM Client Reception Hole with Riva Server-side Combination
Knowledge at the Interface Semantic Innovation and the Shopper Web Experience
Information and the Grid: From Databases to Global Knowledge Communities 
Income Decoupling: New York s Experience Future Directions