• Published on

  • View

  • Download


HUMAN-COMPUTER INTERACTIONHCI RESEARCH METHODSProfessor Bilge MutluComputer Sciences, Psychology, & Industrial and Systems EngineeringUniversity of WisconsinMadisonCS/Psych-770 Human-Computer InteractionMETHODS COMPONENT OVERVIEWMETHODS COMPONENTBuilding a toolbox of research methodsPlanningConducting WritingEMPIRICAL RESEARCHHCI research is founded on empirical principlesDefinition of empirical:Relying on or derived from observation or experiment. American Heritage Dictionary systematic observations of a sample ofIndividual behaviorInteractionsAmong people, between people and objectsUses varying degrees of controlAllows for descriptive or comparative researchMaintaing reliability, validityWill come up repeatedly through the semesterUses specific designsBASIC CONCEPTSWHAT IS A SAMPLE?A subset of a populationThe general population is too large to measureCollecting data from a smaller sample, called samplingMake generalizations on population from the sampleExamples of samplingRandom sampling for controlled experimentsPurposive sampling for representational studiesSnowball sampling used to reach particular groupsConvenience sampling for assignments in this classEXAMPLES OF SAMPLINGAny population of concernSensitive populationsEthical considerationsPrinciples of responsible conduct of research you will learn more in this weeks assignmentSampling biasSelf selection e.g., online and phone-in pollsExperimenter bias e.g., convenience samplingRESEARCH DESIGNDepends on the goals of the investigationGoal-based categoriesGeneralization vs. representationControl-based categoriesLimited control or uncontrolled studiesFully controlled experimentsData-based categoriesQualitative vs. quantitativeAffects design, measurement, data format, analysisGOALSRepresentationIn-depth understanding of phenomena how particular actors affect particular situations under particular circumstancesMainly used for generating theoryUses small samplesGeneralizationTesting hypotheses How findings from a small sample can be generalized to a larger populationMainly used for testing thin slices of theoryUses relatively large samples CONTROLFully controlled experimentsAll aspects but the manipulated aspect are controlled a true experimentField experimentsLimited control in a naturalistic setting e.g., studies in publicObservational studiesNo control on any aspect of the phenomenonSurveys, archival research, cohort study, etc.DATA MEASUREMENTObjective measuresComparing measurements against an objective standard e.g., knowledge testSubjective measuresIndividual and relative evaluation e.g., likingBehavioral measuresMeasuring how people behave instead of reportDATA MEASUREMENTQualitative measuresFly-on-the-wall observationsParticipant observationsOpen-ended interviewsQuantitative measuresQuestionnaires, surveysBiometric measures e.g., eye-trackingTask performanceDATA ANALYSISQuantitative data analysisStatistical methods e.g., counting, t-tests, analysis of variance, time-series analysisQualitative data analysisInterpretations, comparative analysis, modelingQualitative data can also be quantifiedCoding, counting, comparing...EMPHASIS ON WRITINGResearch JournalismGoal is to create a plausible storySelective reportingOpen to interpretationsThe editor can say noResearch JournalismTime span months-old vs. days-old news Rigor systematic process, documentation of observationsExperimentation, interventionGeneralizationTHANKS!Professor Bilge MutluComputer Sciences, Psychology, & Industrial and Systems EngineeringUniversity of WisconsinMadisonCS/Psych-770 Human-Computer Interaction | Fall 2010