Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
DOI:10.3166/LCN.12.4.185-204©2016Lavoisier
FROMANALYSISTOPRESENTATIONInformationvisualizationforreifyingissues
andreenactinginsightsinvisualdataanalysisMARINABOECHAT&TOMMASOVENTURINI
Wediscusstheuseofinformationvisualizationindigitalsociology,(particularlyinControversyMapping),and its role in outlining issues and objects of study through progressive insights. We believe thedifferencesinvisualizationsbetweenanalysisandpresentationarebetterunderstoodaslinkedbyachainof transformations, rather than as two separate and stable levels of representation.Wepropose that,through such chain, two research movements are performed: the reification of issues, related to theconstructionofastableconsensus,andthereenactionofinsights,thatpointstotheroleofvisualizationsas communication tools. We will illustrate such movements and effects by using a few examples ofvisualizationsproducedintheEMAPSresearchproject.
1.Introduction
AsmuchliteratureinScienceandTechnologyStudies(STS)andHistoryofSciencehasestablished,imagessuch as photographs, schematic drawings and graphs have a crucial role in scientific research, either asinstrumentsofinquiry,forsharingmaterialbetweenresearchers,orforadvocatingspecificfindings(Daston&Galison, 2010;Mayer, 2011; Latour, 1985;Offenhuber, 2010; Lynch, 1985).Nevertheless, notmuchof thisliteraturehasbeendealingwiththespecificitiesofinformationvisualizationinscientificactivityandwiththetransformations between visualizations used during inquiry and the visualization employed for thepresentationoffindings.Theseissuesbringforththeroleofvisualizationsforthediscussionsinsideresearchgroupsthatworkwithvisualdataanalysis,andalsotowardsotherdiscussionsasfindingsarepresented.
The discussion we are proposing is related to a fundamental concern of the philosophy of science,representedbythedistinctionbetweenthecontextofdiscoveryandthecontextofjustification.AccordingtoHoyningen-Huene (1987), thisdistinctioncantakeonmanyshapes,but itgenerally refers to thedifferencebetween thediscoveryasanempiricalprocess (and therefore sociologically,psychologicallyandhistoricallysituated), while justification is seen as a set of methods or procedures based on formal logic, to developcriticaltestsforwhathasbeendiscoveredandstreamlinethedescriptionofthediscovery.
Aswediscussthetransformationsofvisualizations,weshouldkeep inmindthatvisualdataanalysis isaprocessofdiscoverythatworkstowardsthepresentationofresultstobuildtheconditionstoitsjustification.Ofcourse,wearenotsayingthatthepresentationoffindingsthroughvisualizationisinitselfthecontextofjustification, but it does set a stage of objects with which justification can be developed. In this sense,visualizationbuildsbridgesbetweenthesetwocontextsinaspecificway:inallthetransformations,wefindvisualdocumentsthatcanhelpoutlineboththeinsightsofthediscoveryandtheelementsofthejustification.
In this paperwe discuss the use of information visualization for visual data analysis in digital sociology,(particularly in Controversy Mapping), and its role in outlining issues and objects of study. Researchersworkingwithdigitalmethods (Rogers, 2013)havebeendeveloping innovativeworkby takingadvantageofthegrowingrichnessofdigitalinscriptionsleftbyhumanactivity.Thosedigitalinscriptionsareseenassourcesof insights,notonlyaboutcyberculture,butaboutsocietyingeneral.Sobyscrapingdatafromsocialmediaandpublicdatabasesand repurposingdata fromvariedsources (Marres&Weltevrede,2013), scholarscanmanagetodeveloprepresentationsofsocialactivityinthemaking.
ControversyMapping isakindofsocialcartographydevelopedfromtheworkofvariousSTSauthors. Itspractitionershave,inthelastfewyears,incorporatedmanytoolsfromdigitalsociology,whileadvancingtheirmaingoal,whichistodescribeandvisuallydeploycontroversies(Venturini,2010).Thatmeansthatthefinalresultsofeachinquiry,composedequallyoftextsandgraphs,willnotaimatestablishingcertainties,butatunfoldingthemeansofadiscussion(Venturinietal.,2015).
186Lescahiersdunumérique–n°4/2016
Our argument derives from the repeatedobservation that visualizations in the exploratory stages differsubstantiallyfromtheonesusedtopresentthefinalfindings:theearlierwilltendtoberawerandclosertothe datasets, while the latter will tend to be more streamlined, displaying aggregate results of analysis.Nevertheless,webelieveitismoreproductivetoconsiderthisdifferenceasachainoftransformations,ratherthanastwoseparatelevelsofrepresentation.Throughsuchchain,tworesearchmovementsareperformed:thereificationofissues,relatedtotheconstructionofastableconsensuswithintheresearchgroup,andthereenactionofinsights,thatpointstotheroleofvisualizationsascommunicationtools.
Unlikewhatisoftenbelieved,thesemovementsdonotnecessarilyleadtosimplervisualizationsinorderto clarify specific findings to awider audience. Visualizations used in final presentationsmay be simple orcomplexdependingontheissuesbeingdemonstratedandontheinquiryitself.Nevertheless,wearguethatthey will be more focused than the ones in the exploratory stages, in the sense of concentrating on theaspects of data that tookpart in the constitutionof the issues.Wewill call these visualisations ‘shallower’visualizations,not inthesenseofbeingsuperficial,butofdisplayingashallowerfielddepth.Theterm“fielddepth” derives fromoptics and is used to describe the interval inwhich objects closer or farther than theexact focuswill have acceptable definition. Field depth is said to be deep (if the focus interval iswide) orshallow (if the focus interval isnarrow).So the term,asusedhere forvisualizations,points toaprocessoffiltering,definitionandaggregation,ratherthanofmeresimplification.
We will illustrate our two movements by using a few examples of visualizations produced in EMAPS(http://www.emapsproject.com/),aEuropeanresearchproject.Between2010and2014,scholars,designersand developers from different research institutions got together with different stakeholders (called ‘issuespecialists’) to experiment with visual data analysis and develop visual representations of the debates onageingandclimatechangeadaptation.
2.Visualizationsbetweendiscoveryandjustification
Asclaimedbynumerousscholars,imagesarecrucialforscientificactivity,fromthemanydevicesusedforexploration at the beginning of the inquiry towards the standardized and measured representations thatoutlineepistemicobjectsanddemonstratescientificdiscoveries.Daston&Galison(2010)extensivelydescribethe role of scientific atlases in the development of a collective empiricism and of different scientificethosfromtheeighteenthcenturyon.Serres&Farouki(1999)seescientificimagesasculturalobjects,andpointtotheir presence in everyday life and to their role in rediscovering wonder in the world. Offenhuber (2010)discussestherhetoricalpowerofvisualizations,andtheircapacityforengenderingnarratives.Regardingthesocial sciences more specifically, Healy and Moody (2013) describe a history of the use of informationvisualizationinsuchfields,relatedtoissuessuchasreliabilityandtheeaseforsharingcodeanddatabetweenresearchersandwiderpublics.
Inanarticlecalled“Lesvuesde l’esprit”,Latour(1985)describestheroleof images,drawingsandvisualrecords for constituting “immutable mobiles”, i.e. standardized and stabilized objects that could berecombined and transported to different contexts. Later on, in Pandora’s hope (1999), he describes theprocessinwhichsuchobjectsaredeveloped,inaseriesoftransformationsbetweenaninitialobjectofstudy,thatiscomplexandmessy,toprogressivelycompatible,standardizedandreliableelements.Lynch(1985)hadalreadydescribedaprocessof thesamesort, regardingmechanicallyproduced(orphotographic) images inbiologicalsciences:accordingtohimsuchprocessesinvolvesstepsofmathematizationandschematizationofimages,untiltheybecomestreamlinedandshareabledocuments.
All theseworks point to themultiple transformations that scientific images undergo during the inquiry.These transformationsareparticularlyevident indigitalvisualization.Unlikemechanicallyproduced images,digital visualizations have no indicial relationwith the object of study. They are derived not from a directmanipulationoftheirobject,butfromamanipulationonthedigitalinformationcollectedonit.Becausethisdouble mediation (that of digitalization and that of the transformations operated in the computer), the
Fromanalysistopresentation187
spatiality of digital visualization is even less constrained by the resemblance to its object andmay changedrasticallyasresearchersexperimentwithdifferentmethodsofvisualization.
Visualizations are, atonce, tools for exploring a contextof inscriptions and fordisplaying configurationsthatadvanceaspecificdescriptionofthings.Becauseofthisdoublefunction,theycanplaydifferentrolesindataanalysis.Atoneend,theyworkasinstrumentsfordiscovery,promptinginsightsandsuggestingfindings,whileattheotherendtheyconvey‘hardened’factsandtheirdemonstration(Latour,1999).
The transformation of information visualizations seems, to some extent, to follow the distinction(customary in thephilosophyof science)between thecontextofdiscoveryand thecontextof justification.According toHoyningen-Huene (1987),diverse interpretationsof thisdistinctionhavebeenput forwardbyauthorssuchasPopper,KuhnandFeyerabend,whodepicteddiscoveryandjustificationas:
(a) twoprocessesthatfolloweachother,withthefirstbeingtheconditionforthesecond;
(b) twocounterparts,withjustificationfunctioningasacriticaltesttoreconstructdiscovery;(c) twodifferentmethods,withdiscoverybeing“anempiricalenterprise, (...) [that]may involvehistorical,
psychological and sociological reasoning” (p.505), while justification (or critical testing) beingfundamentallylogical;
(d) the object of different disciplines,with philosophy of science addressing the logic of justification, andhistory,sociologyandpsychologystudyingtheempiricalprocessesofdiscovery;
(e) thedifferentresultsobtainedbyaskingquestionsaboutdiscoveryoraboutjustification.
Thoughthejustification/discoverydistinctioncapturesinbroadtermsthedirectionofthetransformationsundergonebyscientificimages,themodelthatitproposesisfartooabstracttodescribetheactualworkofinformationvisualization.Suchwork,weargue,bridgesthegapbetweendiscoveryandjustificationinaveryspecificway,andcandisplayevidencesofhowthetwocontextsinteract.
Certain kindsof inquirypromotea clearerdistinctionbetweenexplorationandpublication, analysis andpresentation.Inatraditionaldemographicstudy,forexample,thefinalfindingsareexpectedtobeashardaspossible, so the corresponding visualization must be concise and clear– no matter how many tests wereconductedduring analysis, or howmuchexploration it took to reach those conclusions. Yet, inmostothercases,thetransformationof informationvisualizationwillnotgoso linearlyfromexplorationtopublication.Controversy mapping, in particular, will tend to emphasize the many translations between analysis andpresentation,fortworeasons:first,becauseitadoptsaparticipativeapproachinwhichvisualizationsshouldremainopeninordertoassurethecommunicationamongparticipantscomingfromvariedbackgrounds;andsecond,becauseitsverygoalistoencouragethedebateratherthantoreachcertainties.
3.Visualizationsincontroversymapping
Controversy Mapping is a method of social research that uses visual analysis to produce maps of theassemblies that actors form around disputed issues. It derives from a long tradition of dispute analysis inscienceand technologystudiesandactor-network theoryandhasbeendeveloped in itsdigital format themédialabatSciencesPoParisandattheDigitalMethodsInitiative(DMI),attheUniversityofAmsterdam.
Incontroversymapping,visual translation ispartof the inquiry itself,partof refinementof theresearchquestionsandof theobjectsof study.As theworkprogresses,newobjects likecategoriesandclustersareoutlined and it is possible to elaborate other ways of treating the data and visualizing it, according toquestionsthatbecomeprogressivelymorepreciseandarematchedwithmoredefinedobjects.Eachversionincorporatesandmakesvisiblemoreandmoreinterpretationandanalysis.Therefore,inthedevelopmentofinquiries inEMAPS,newobjectsareproducedwhosespatialitydoesnotnecessarilymapback to the initialvisualization,andthedatacanbefilteredandconvertedintootherstructures.
AnexampleofthisprocessisdisplayedinaworkcarriedoutbytheSciencesPomédialabonthemappingof the scientific literature related toCO2. Figure1 illustrates theprotocol behind this typeof research and
188Lescahiersdunumérique–n°4/2016
figure2providesastreamlined,yetstillcomplex,versionofthenetworkfortheyearsof1960to1969.Theimagewas produced by querying the ISIWeb of Science for the keywords “carbon dioxide” or “CO2” andextractingallthereferencescitedintheresultingbibliographicalnotices.Thesereferencesareusedtobuildan initial networkof reference co-occurrence.Aswe can see in the illustration, this firstnetwork is almostillegibleanddoesnotprovidemuchinformation.Asecondversioniscreatedbyusingaforcevectoralgorithmthatmakesclustersevident(Jacomyetal.,2014).Afterthat,anotherlayerofdataisproducedbyextractingmetadata from the bibliographical notices (authors, institutions, keywords, disciplinary categories…): thisinformationisdisplayedasnewnodesconnectedtothereferences(akeyword,forexample,isconnectedtoagroup of reference if they appear in the same bibliographical notice). In this study it is clear howtransformationsinvisualizationsareintertwinedwithtransformationsindata.
Figure1.CO2LandscapefromISI-WoS,MethodDiagram,byVenturini&DePryck
Figure2.ScientometricmapdesCO2,byVenturini&DePryck
4.VisualizationintheEMAPSProject
EMAPS is a European collaborative project completed in 2014 and developed by a consortium of sixEuropeanresearchcenters (SciencesPomédialab,DMI fromUvA,YoungFoundation,PolimiDensityDesignLab,BarcelonaMediaandDortmundInstituteofSpatialPlanning).Itsmainresultwasasetofmapsaboutthecontroversiesonclimatechangeadaptationthataimedatmobilizingdigitaldatatoequippublicdebate(seethe project blog at http://www.emapsproject.com and the final results at http://www.climaps.eu). EMAPSmixed the different research traditions of its partners becoming a ground for experiments in digital socialresearchandinformationvisualization.
DuringtheEMAPSdatasprints(Venturinietal.,2016)wewereabletofollowindetailthetransformationsof visualizations during data exploration. In the next pages, we will discuss a particular chain oftransformations started in theAmsterdam sprint, inMarch 2014. The sprint gathered scholars, developersanddesigners from theparticipant centers in theUniversity ofAmsterdam for fivedays of intensivework.Participantswereorganizedinfivegroups,eachworkingwithdifferentdatasetstoexploredifferentresearchquestionsaroundthethemeofclimatechangeadaptation.Group4(namedUsesandUsersofVulnerabilityIndexes),carriedouttwoprojects:thefirststudyingtheusesofvulnerabilityindexes;thesecondexploringtheextenttowhichflowsofadaptationfundsarerelatedtovulnerabilityranking.
Fromanalysistopresentation189
Forthesecondproject,thevisualizationswereintendedtodiscusstwomainquestion:– Arethecountriesconsideredmostvulnerablealsotheoneswhoreceivethemostadaptationfunds?– Do the countries considered vulnerable by some indices receive more funds than those considered
vulnerablebyotherindices?The studied indexes were: theDARA Climate VulnerabilityMonitor1,Germanwatch’s Climate Risk Index
20142 andMaplecroft’s Climate Change Vulnerability Index3. TheUN Human Development Index4 was alsoused as the researchers identified in the debates a perceived link between climate vulnerability and lowerhumandevelopment.Adatasetontheallocationoffundsbycountrywasavailable(thanktothecollaborationof climatefundsupdate.org) for each of themajor international funds: theAdaptation Fund (AF); the LeastDeveloped Countries Fund (LDCF); the Special Climate Change Fund (SCCF); and the Pilot Programme forClimateResilience(PPCR).
Figure3.AdaptationaidperFund-GermanwatchIndex.Oneofthevisualizationsintheseriesthatdisplayedtheallocationsofeachfundacrosscountries,orderedbyvulnerabilitytoclimatechange,accordingtotheGermanwatchIndex.Theallocationsarerepresentedbytheirvalueindollars.Source:EMAPSarchives
Twokindsofbubblegraphsweretriedout.Thefirst(figure3)positioncountries-bubblesinascatterplotaccordingtotheirlevelofvulnerability(X-axis)andtheallocationsofthefourfunds(Y-axis).Theothergraph(figure4) was a long list of countries sorted by vulnerability, with circles proportional to the amount offundingreceivedbyeachfund.Thisseconddiagramhadseveraldesignproblems:itdidnotprovideageneralviewofthedatapoints(scrollwasneeded)anditdisplayedlessdimensionsthanthefirst,failingtoshowtheactualamountsoffundsthatwereallocatedforeachcountry.However,itdisplayedtheamountsintermsofthe percentages of the total budget each fund allocated, thus comparing the priorities of each fund tovulnerability according to each index. Another difference is that it was flatter: instead of allowingsuperpositionsthegraphspreadalldatapointssidebyside.Thisallowedfordifferentconsiderations,moregeared towardsdiscussing criteria of funders and the relevanceof the indexes, andnot the gross financialresult.
1.http://daraint.org/climate-vulnerability-monitor/climate-vulnerability-monitor-2012/monitor/2.https://germanwatch.org/en/cri3.http://maplecroft.com/themes/cc/4.http://hdr.undp.org/en/content/human-development-index-hdi
190Lescahiersdunumérique–n°4/2016
Figure4.AviewoftheHTMLvisualizationthatdisplayedproportionalallocationsofeachfundforeachcountry,orderedaccordingtotheGermanwatchIndex.
Source:EMAPSArchives
Fromanalysistopresentation191
Figure5.MultilateralAdaptationFundingandVulnerabilityIndexes.Theinteractiveversion,whereonecanchooseanindexforthexaxisandthefundstobeincludedinthemapping,thuschangingthesizeofthecirclesproportionally.Source:http://climaps.eu/#!/map/multilateral-adaptation-funding-and-vulnerability-indexes
Figure6.Anotherviewfromthepreviousvisualization,displayingonlytheLDCFundinthecoloredareas,withthegraycirclesindicatingthetotalfundsallocatedtoeachcountry.Source:
http://climaps.eu/#!/map/multilateral-adaptation-funding-and-vulnerability-indexes
Figure7.MultilateralAdaptationFundingandVulnerabilityIndices-Matrix.Thebarsforeachcountryareorderedverticallyaccordingtoeachindex,andthecolordensityisproportionaltotheamountallocated.
Source:http://climaps.eu/#!/map/multilateral-adaptation-funding-and-vulnerability-indices-matrix
192Lescahiersdunumérique–n°4/2016
InthefinalresultspublishedinClimaps.eu,thetwoapproachesarepresent:figures5and6showdifferentviewsofaninteractiveversionofthegraphinfigure3.Nowthecolorsofthebubblesareusedtoreinforcethegrowingvulnerability.Infigure7weseeamuchclearerandcondensedversionoftheinitialHTMLgraphinfigure4.Bothfinalversionspresentaricherdescriptionsoftheissueandarenotnecessarilysimplerthantheinitialones.
IntheOxfordsprint,thattookplaceinApril2014,participantsweregatheredinfourgroups.TwoofthemusedvariationsoftheHTMLbubblegraphinordertodisplaysquarematrices.Group3,thatsoughttoprofileadaptation practices, tuned this diagram into a large graph for initial exploration and comparison ofadaptationsprojectsalongmanytopics(weseeapartialreproductionofthisgraphinfigure9).
Group1wasalsointerestedinprofilingprojects,butforunderstandingwhichhazardsweremorerelatedto vulnerability. So the same structure was used, but with data coming from two different sources: thedatabasesfromUNDP5andci:Grasp6.Bothofthemareavailableinwebsitesofpublicaccess,inseriesofwebpagesthatofferafullviewofeachprojectbutdonotallowcomparisons.TheEMAPSvisualizationcombinesthetwodatasetsandfacilitatethecomparisonamongmorethanthreehundreditems.
Diagramsthatfacilitateaninitialappreciationoflargeamountsofrecords,areoftenthestartingpointformanyfollowingtransformations.Thegrid infigure10, forexample,wasusedtocomparethecasesof IndiaandBangladesh.Thiscomparisongeneratedthegraphsinfigures10and11,andwasrepresentedinamorecomplex interactive visualization in Climaps.eu (figure 12). The final interactive visualization replaces thebubbles with bars with the advantage of showing the proportion of each value to the maximum, aidingcomparison.Nevertheless, thegeneral table-likestructure ismaintained: insteadofamatrixof thebubbles(figure4),wenowhaveamatrixofbars.
We identify threemainmovements of transformation in this case: first, the display of the first bubblegraph is refined into amore streamlined presentation; second, the structure of the big grid is polished inordertofunctionasagenerictoolforotherresearchquestions;third,theanalysisofthedatadrivesnewdatatreatmentsandtheproductionofmoreadvancedvisualizations(seeaschematicsummaryinfigure12).Inthesecondmovement, thebubblegraphwasturned intoasquarematrix: this timeextravisualelementswereaddedtogenerateageneraltoolforintegratingdifferentdatasetsandofferinganinitialdataexploration.Inthe thirdmovement, the structure is used to reconcile different datasets so that researchers could have amore complete view, and afterwards specific data points and details identified in the grid are displayed inmorelimitedvisualizations.
5.Shallowness,reificationandreenaction
It should be clear by now that, across work groups and data sprints, from analysis to presentation,simplificationisnotanadequatewayofcharacterizingtheeffectsofvisualtransformations.
Whileexaminingthesetransformations,weinitiallycomeacrosstwointerpretationsoftheideaofvisualsimplification.First,wecouldthinkofsimplificationasaprocess inthechainoftransformations inscientificimages (Lynch,1985). Itcanbeunderstoodasaprogressiveschematizationthatcondensatesrelevantcuesand thus visually simplifies an initial messy object. Second, we could think of simplification in relation tocommunication,aspartof theeffortofmakingvisualizationsmoreaccessible.This second interpretation isoften present in the literature of information design. Tufte (1983), for example, talks about improving thedatainkratioofgraphics,arguingfortheremovalofallthevisualinformationthatdoesnotdisplaydataandisthereforejustdecorative.
Thehtmlbubblegraph,despiteitsmanydesignproblems,diddisplayveryclearlythe(lackof)correlationbetween the priorities of funders and the vulnerability indexes. We believe this is related to its very flatpresentation(withthebubblesdistributedasifinatable,withnosuperpositionandnodepthinpresentation)
5.http://adaptation-undp.org/6.http://pik-potsdam.de/cigrasp-2/
Fromanalysistopresentation193
andanormalizeddataset,wheredifferences in thesizesof totalbudgetswereclearedby treating themaspercentages.Sincetheresearchquestionwasgearedtowardsunderstandingtheprioritiesofeachfund,andcomparing them to the priorities suggested by vulnerability indexes, displaying amounts in dollars wouldintroduceunnecessarydetail(consideringtheverydifferentallocationsofthedifferentfunds).Inthissense,theinitialbubblegraphisshallowerthanthebettervisuallydesignedvisualizationinfigure3,because,ontheone hand, it is more limited and more focused, and, on the other hand, it is more regular from a visualstandpoint.Weconsider the final streamlinedversion in figure7 tobeevenshallower,because itdoesnotdepend on navigation or scrolling to display all the data for all indexes: it organizes them side by side tofacilitatecomparison.
Indeed,byprogressivelyframingissuesinvisualizationswetendtoaggregatemanyelements inmoreorlesscohesiveentities.Reconstructingtheinsightsoftheenquiry,wetendtodisplaytheseinsightsasentitiesthemselves, building highlights and clearing out elements that are not directly connected to thedemonstration.Movingfromanalysistopresentationdemandsthehardeningoftheobjectsofstudy,aswellas the progressive outlining of narratives and rhetorical strategies. This movement does entail somesimplification. Yet, this simplification should not be mistaken for an objective in itself or, even worse, anecessary didactic effort for the audienceof the final presentation. In fact, simplificationhappens as a by-product of defining objects and developing the inquiry. Instead of thinking in terms of simplification, weproposetheideaofprogressivelyshallowervisualizations, inthesensethattheyflattenthedata landscape,not by avoiding complexity, but by reducing the depth of focus . At each step, the visualizations do notbecomesimpleranddonotnecessarilydisplay less information,butdo indeedbecomemorecoherentanddisplayelementsmoreneatly.
Figure8.Thebiggrid,asquarematrixtointegratedatasetsfromdifferentprofilingsourcesandtohelpidentifyingpatternsbetweenmanyprojects.Source:EMAPSArchives
194Lescahiersdunumérique–n°4/2016
Insteadofagrowingsimplification,weobservetwomoresubtlermovements:thereificationofissuesandthereenactionofinsights.Inthefirstmovementweareusingtheterm“issues”(andnot‘scientificobjects’or‘facts’)tohighlighttheroleofvisualizationasatoolforacademicandpublicdebate.Wealsodonotwanttolosesightoftheideathatscientificobjectsandfactsareproducedthroughresearchwork(Latour,1999),andthatvisualizationsgivesvisualcluestoadvanceofsuchprocesses.This firstmovement iscomplementedbyanother movement that we call the reenaction of insights. This latter is related to the concerns aboutcommunication, which entail didacticism and rhetoric (Offenhuber, 2010). The insights that happen in theanalysis stageshave tobe re-produced (or demonstrated) inpresentation, therebymaintainingpart of theexploratoryperspectiveeveninthefinalresults.
Figures9and10(fromtop):visualizationsdraftedtohighlightaspectsoftheprofilesofIndiaandBangladesh,thatservedforthecommunicationbetweenresearchers.First,thecomparisononvulnerabilityranking,andsecond,thecomparisonofnumberofprojectsbyfundingsource.Thesegraphstakeonestepfurtherfromtheinitialgrid,selectingonlysomeofthedatapointsconsideredtobethemostrelevant,afterthegeneralviewthegridprovided.Source:EMAPSarchives
Thechainof transformations fromanalysis topresentation follows, tosomeextent, thedevelopmentoftheinquiry.Inmostcases,thisdevelopmentisnotlinear:allassumptionswillbequestionedandsomewillbeabandoned;explorationswill leadtodead-ends;visualanalysiswill sometimeonlydisplaythe limitationsorthevicesofdatabase;all researchquestionswillhave tobeadjustedandmanywillbediscarded.Likewise,visualizationsarealtered,cutout,madeagain,becomethebasetoothers,discarded,replaced,refined.Muchlikewhatweseeintheschemapresentedinfigure12.Visualizationscomment,detailandcontextualizeothervisualizations, andmay also build over the work done in previous ones,making these last disposable andoutdated.
Fromanalysistopresentation195
Figure11.OneofthevisualizationsfinallypublishedonthesiteClimaps.eu:atoolwheretheusercanverifythefindingsexposedinthetextorsearchformeaningfulcomparisons.Thevisualizationproposesacomparison
betweenIndiaandBangladesh,butthereisalsothepossibilityoftakingabroaderviewandseeingthedataonalltheprojectsfromcountries.Source:http://climaps.eu/#!/map/what-is-an-adaptation-project-ii
Nevertheless,itisbyopeningandthreadingthesepathsthatclassificationsarecreated,alliancesoutlined,clustersidentified,patternsrevealed.Alltheseitemsbecomevisibleentities,thingsthatscholarscanpointto,compare,discussandchallenge.This iswhatwecallthereificationof issues,and it ispartoftheexchangesinsideresearchgroups.Itisthelabornecessarytoprogressively(thoughnotlinearly)framingtheissuesanddevelopingthediscursiveandvisualtoolstoaddressthem.
Theothermovementofthisprocess,thereenactionof insights, ismorecloselyconnectedtotheroleofvisualizationsastoolsforengagingwideraudiences.Offenhuberpointstoarhetoricalmaneuverthatisusedwhenresearchersdemonstrateresultsofvisualdataanalysis:
“...afterpuzzlingtheaudiencewithacomplexvisualization,thepresenterselects,seeminglyarbitrarily,a single data point and connects it to a story, an anecdote that unlocks the principle of the wholerepresentation.Isuspectthissingledatapointisseldomasarbitraryasitmightseem,infactthewholevisualizationmightbedesignedtohighlightthissinglepoint–arhetoricaldeviceallowingtheaudiencetoreproducethediscoveryofmeaninginthedata.”(Offenhuber,2010,p.370)
Figure12.AsummaryofthetransformationsobservedinEMAPS
196Lescahiersdunumérique–n°4/2016
This strategy, called “visual anecdote”, convincesby reproducing the transformations carriedoutduringanalysis.This relates towhatLatour (1999)calls thechainof referenceofscientificartifacts,whereateachstep of transformation it is important to keep the link to the previous step in case results need to bereproduced. Locating the reenaction of insights only in the final results, however, would be too narrow.Research projects gather different people, with different backgrounds and the visualizations also serve tomakesurethe“everyoneisonthesamepage”.Throughoutthewholeanalysis,thereenactionofinsightsisthereforepartof thediscussionthatsupports theenquiry. Ingeneral,ourcaseprovidesastrongargumentagainst the separation of discovery and justification, at least in information visualization. If EMAPS wasproductiveingeneratingfindingsandcommunicatingthemitisbecauseiteffectivelycombinedthereificationofissuesandthereenactionofinsights:thefinalmapscommunicatebydemonstrating.
References
DastonL.,GalisonP.Objectivity.ZoneBooks:Brooklyn,2007.
HealyK.,MoodyJ.(2014).Datavisualizationinsociology.Annualreviewofsociology,vol.40,p.105-128.
Hoyningen-Huene P. (1987). Context of discovery and context of justification. Studies in History and Philosophy ofScience,vol.18,n°4,p.500-515.
JacomyM.,VenturiniT.,HeymannS.,&BastianM.(2014).ForceAtlas2,aContinuousGraphLayoutAlgorithmforHandyNetworkVisualizationDesignedfortheGephiSoftware.PloSOne,9(6),e98679.
LatourB.(1985).Vuesdel’esprit,Culture&Techniquen°14.http://documents.irevues.inist.fr/handle/2042/29797
LatourB.(1999).Pandora'shope:essaysontherealityofsciencestudies.Harvard,HarvardUniversityPress.
LynchM.(1985).Leretineexteriorisé.Culture&Technique,n°14.http://documents.irevues.inist.fr/handle/2042/29797.
MarresN.,WeltevredeE.(2013).ScrapingtheSocial?Issuesinlivesocialresearch.JournalofCulturalEconomy,6(3),pp.313-335.
MayerK.(2011).Scientificimages?Howtouching!.Science,Technology&InnovationStudies,7(1),p.29-45.
OffenhuberD.2010.VisualAnecdote.Leonardo,43(4),p.367-374,Available:http://dspace.mit.edu/handle/1721.1/60338(Access:01/05/2014)
RogersR.(2013).Theendofthevirtual:digitalmethods.Cambridge:MITPress.
SerresM.,FaroukiN.(1999).PaysagesdesSciences.Paris,ÉditionsLePommier-Fayard,
TufteE.(1983).TheVisualDisplayofQuantitativeInformation.Cheshire,GraphicsPress.
VenturiniT.etal.(2015).DesigningControversiesandtheirPublics,DesignIssues,n°31,vol.3.
VenturiniT.,MunkA.,&MeunierA. (2016).Data-Sprint:aPublicApproachtoDigitalResearch. (C.Lury,P.Clough,M.Michael,R.Fensham,S.Lammes,A.Last,&E.Uprichard,Eds.)InterdisciplinaryResearchMethods(forthcoming).
VenturiniT.(2012).BuildingonFaults:HowtoRepresentControversieswithDigitalMethods.InPublicUnderstandingofScience,n°21,vol.7.