good morning everyoneum or good afternoon or eveningdepending on where you are dialing infrom todaymy name is mai sisla and i’m deputydirector of the aspen tech policy hubfor those of you who are new to us theaspen tech policy hubis a bay area based policy incubatortraining technologists on how to engagewithand improve policy i’m thrilled towelcome you to the second of threewebinarsreleasing products from our kovid 19challenge grant seriesuh we launched our kovit 19 challengegrant series in early april when we sawthat there wasa potential for technologists tocontribute to mitigate some of thenegative effects of the pandemicwe funded six projects in total withthese challenge grants and today’swebinar is presenting two of those sixprojects um the thread connecting thesetwo projects todayare those that help us better measureand understand risks that have come upduring the pandemicthis includes risks from virustransmission itself as well as risksfrom other consequences includingmisinformationand today’s projects are aimed atbuilding informational understanding andresponse in these pandemic conditionsbefore we get started i wanted to give abrief run of showthe two grant teams will present theirprojects each for about 10 minutesthen we’re thrilled and excited to havekeynote speaker and technologystrategist rachel coldy cutformer ceo of dot everyone to respond toand help put these projects into contexta couple of logistical announcements aswell there’s a qa box at the bottom of your zoom screenwe’ll have plenty of time for questionsat the end so if you just put yourquestions in there we’ll be sureto put them in the queue for the end andsecond the grantees are releasing theirprojects and products liveso if you check the chat box the linksto the products and further informationwill be providedum so i’m thrilled to introduce ourfirst set of granteestommy shane and pedro noel they come tous from first draft which is anorganization that aims to protectcommunities from harmful misinformationtheir project the data deficitsdashboard focuses on building a toolthat maps out thedemand and supply of credible covid 19informationin order to better track misinformationand develop effective policy solutionsso in order to learn more about theirprojects we’ll turn it to themover to you tommy and pedrohello everyone um hopefully you can seemy slidesand um yeah i’ll begin so thanks so muchtouh the aspen institute for hosting thiswebinar today um and for everyone herewho’s joining usum i can talk a bit about projectsaround data deficitsum i’ll talk a bit about what that meansandi’ll be passing over to my colleaguepedro who’ll be talking aboutthe dashboard that we built to uh tomonitor themso um i’m tommy shane i work at firstdraft whichis a company that tackles misinformationgloballyand i’m ahead of the impact on policythereand um what i want to say today is abouta content called data deficitswhich is related to misinformation so inorder toexplain what data deficits are i want toroll back theuh the clock to the 25th of januaryearlier this year when questions beganto emerge about howit was possible that the wordcoronavirus was on the back ofdettol bottles now for those who don’tknow what dettol is it’s a it’s anantibacterial sprayso these are the kinds of uh posts thatuh emergedin relation to the to the labels becauseif you can see here on the slidethe word coronavirus featured on thesebottles even though they were from uh2019 a long time before the coronavirusstruckum now these posts might seem kind ofharmless butwhat they did is they they contributedto a narrativethat people knew something about koreanadvice people knew what it was thatmaybe there’s something that people werekeeping from us that they’re not tellingusthis is exactly the kind of rumor andspeculation that can driveconspiracy theories now more importantthan thatwe also know that um some peoplethought that maybe if you could justglug down a bottle of thisyou might cure yourself and obviouslythese are very harmful and dangerousspeculations and questions that peopleinevitably ask after acrisis strikes now it wasn’t untilthree weeks later on the 17th offebruary thefull fact that fat checker in the uk wasableto fact-check this claim and to providecredible informationto say that there is no evidence thatdetail can helpso what was this like what was thisperiod of timethis situation people wanted to knowabout death so it was kind of veryhigh vulnerability around asking thisquestion and getting the wrong kind ofinformationand then a delay the the delay beforecredible informationcame onto the scene to answer thesequestions well this was a question thatwe wanted to to tackle but it wasn’t areally clearconceptual framework or a term or anidea to really tackle thisso what we wanted to do was to thinkaboutthe empademic and incredible informationin terms of demand and supply so on theone hand you have demandsso what do people want to know aboutright now in the case of data bottles itwascan it kill coronavirus did they knowabout itbefore anyone else did so that’s thedemand for informationand on the other side we have suppliesso the supply of credible informationin this case in the form of a fat checkfrom uh full factum and what we saw here was that we hada high level of demand to start withum but we had low levels of supply ofcredible informationand there was a disparity there and whatwe wanted to do was to try and targetthat problemso we have high interest below supplyand we this is what we’re calling adeficit um high levels of demand and lowlevels of supplynow this might all sound verytheoretical and abstract and like okaygreatnow you’re just calling you’ve got allthese terms you’ve got these words butwhy is this actually helpful for anyoneand the reason is that when we thinkabout demand and supply we create thisnew framework to think aboutdifferent signals some data points thatwe can monitor in real timeto start to understand where thesevulnerabilities these risks are emergingso let’s take a look at demand so demandiswhat do people want to know about rightnow so now we can start to ideatewhat signals out there on the internetcan help usum answer this question what do peopleknow about right nowso we’ve got what are people searchingfor on google and that can tell ussomething about what people want to findout aboutpeople searching for on on social mediawhat people are trying to learn about onwikipediawhat do people really care about enoughto petition the government aboutuh what pages are people viewing ongovernment websites whatsocial posts are going viral moreobviously what narratives are becomingpopularin in the press and on online there’s awhole set of metrics that we can startto monitor in real timeand if we look to supply uh you knowwhat credible information are peopleaccessingwe can again look at a whole differentset of metrics sowhat fat checks are people accessingwhat rate research papersare out there um what news stories arebeing published what’s the who and othermedical bodies saying what governmentstatements sayingand so when we have this new frameworkwe can get these all these differentmetricsfor the demand side and the supply sidewe can start to bring them togetherand then say okay well look where are wefinding instances of really high demandand really low supply and that is wherewe need to actso this was the real question thatanimated our project where are thesedeficitsand how can we try and monitor them inreal timeand in order to put this wholeconceptual framework into practicewe wanted to build a prototype dashboardand my colleague pedro will explain abit about how we actuallydid thatthank you tommy hi everyone thanks againfor for joining usuh so we decided to uhbuild this dashboard in order to helppeople to visualizeuh these deficits and depths ofengagement and the distribution ofdebugscan you pass the next slide tommy pleaseso we decided to use a limited array ofindicators to calculate this demand andsupply during this fastsprint so for demand wedecided to use indicator related togoogle searches so for that we used theindicators provided by google trendsuh as also we used indicators based ontop articles on wikipediafiltered by language for supplywe used fact checks as an indicator ofcrediblecoverage on coffee 19.and next slide please uh so in total wegathereduh 9722debunks uh in order to provide usindicators regarding supplythese debunks were extracted from threesources umifcn which is internationalfact-checking networkuh those organizations fill in the claimreviewand uh full facts uh debunks from factwhich we extracted from thefacts native api so it’s arounduh the data set comprises around seventhousand debunks from an fcn2000 from claim review and around 80fromfull fact now we wanted tocategorize these debunks in terms oftopicsso we worked with universe of sheffieldin order to use their machine andmachine learning modelto categorize and assign topics to thiseach of these debugs the typology weuse it was based on the typologyproposed by the writers instituteof the university of oxford in a reportthey published it onapril uh in april this yearyou can go to the yeah uhin order to gather uh indicators fordemand so we use it the google trendsapiuh so basically we calculated theaverageuh of interest on coronavirus on theterm coronavirusuh in the past 48 hours so as i showedlaterthere’s a live display of this uhindicators in the dashboard and alsoa display related to the to the currentand actualindicator for these searches related tocoronavirusas for the top articles from wikipediawe used the nativeum wiki wikimedia api in order to gatherthe top 10[Music]most read articles on wikipediauh the dashboard itself is comprised byfoursections uh the first sectionis is a map where you can have ageographic visualizationon where these data deficits are theengagement deficits areum and uh indicators about the searchintereststhe total engagements these debunksreceive it the average engagementum of for for these debunks andthe final total number of debunksassigned to the to these countriesuh on such the second section is abreakdown per countryuh of top searches onon on google extracted from googletrends thistop most read the most read wikipediapagesin the native language of this countryand alsoum a breakdown of the top fact checksthat’s circulated in this countrythird section is uh about whereuh which fact checks resonated most soyou can have a breakdown of the factchecks the organization publishing thesefact these checks the topic of this uhpieces and the engagement these piecesreceiveduh fourth section uh is aboutuh the breakdown of the countries soit’s basically a display of the numberof fact checks that circulate in thiscountryuh the the indicator of data deficitof this country total engagement ofthese debunks the engagement average andtheengagement deficits all this again isfilterableby by topicum yeahso if you want to see and access what wedidyou just go to the to this url which isdatadefsis.firstraftnews.orgthere also you’ll be able to see the thewhole report we publish it with theum with the explaining like the thethe methodology we use it in in detailso this isthe urlso just to finalize uh just like a fewremarks on what we learnedduring this process first that managingsuch a high volume of spice is hard notonlydue to the logistics of managing suchamount of datawe had like a hard time deduplicatingdebunks we hadissues with the annotationof this debunks performed by theorganizations byclaim review by the fact checkersthemselvesalso we faced issues related to apilimits uhfor example so in order when we weretrying to extract metrics from this umdebunks um it was very difficult to dothis in a quick waybecause uh most of the apis they havelimit time limitswhich somehow sometimes block or makethe processmore difficult uhalso we don’t have the right signalsfrom social media platformsbecause as we asas we stated in previous piecespublished byfirst draft the search engine socialmedia platforms they’re becoming searchengines as welland there is no actual metrics orwe don’t have access to the what peopleare searching in these platforms rightlike the indicators provided by that bythe platformsare very central on interactions and notspecifically in what people aresearching therefinally we need to better understand howthe information provided by googletrends so althoughyou can use for example some operatorsto perform queries on google trendsgoogle trends does not provide a fullyboolean friendlymethod of search so we we facedissues on querying for specific terms wewere obligated to queryfor the coronal virus as a general termto evaluateall this so these are the few remarksaboutwhat we learned in this process that’sitthank you very much i hope to be able toreply to your questionsvery soon and back to you minewonderful thanks so much tommy and pedroum between the upcoming elections herein the usand some of the misinformation we’veseen already from the coven pandemicfiguring out how to measure sort of anebulousconcept like misinformation or datadeficits is is a key step intofiguring out how to combat it so wereally thank you for your workum just a reminder for the audiencemembers if you have any questions toplease submit them in the q a boxbelow um and also check the chatfunctionfor links to the actual dashboard thatpedro andtommy just discussed and some of theirreportsum next i’d like to introduce our secondgrantee of todayum pietro michellucci of the humancomputation instituteum the human computation institute is anon-profit innovation centerdedicated to the betterment of societythrough novel methodsthat leverage the complementarycomplementary strengths of networkedhumans and machinespietro and his team are developing anapp informed by epidemiology andbehavioral scienceto help people reduce covid19transmission by making informeddecisions about where and when to runerrands and we’ll be learning more alittle bit a little bit more about thisproject from pietro so over to yougreat thank you my and uh frankly ithinkyou said everything i was going to sayso we can just stopright here um so umas uh as mai said uh my name is pietromicheluccii lead the human computation institutewhere we uhdevelop uh and build crowd-poweredsystems to tackle societal problemsso for example we created an online gamecalled stall catchersthat now has 30 000 players speeding upcornell university’s alzheimer’sresearchso when coveted 19 emerged we started tobrainstorm about whether we couldsomehow apply these same methods toreducing the impacts of the pandemicand we came up with an exciting idea umso we were thinking where uh do peoplego when they’re social distancing wheredo they still have to goso to the store to run errands so wehave this idea that when lots of peopleall go to the same storethe infection risk is higher and if wecould evenly spread people out acrossstoresmaybe we could reduce virus transmissionso we did some napkin math to see if thenumbers work out so if you take ninepeopleput them all in the same store eachperson is exposed to eight other peopleeach of whom could already be infectedwith covid so that’s 72 different waysthat infection could spreadbut what if you had three stores and youcould distribute those same nine peopleso that there were only three peopleat each store then each person would beexposed to only two other peoplewhich adds up to six ways infectioncould spread in each storeor 18 total infection opportunities sospreading people out evenly resulted infour timesfewer infection opportunities the napkinmath confirmed our intuition but thenext question washow do we get people to actually do thatand we realized that given the optionsome people might drive a little fartherto reduce their infection risk by goingto a less crowded storeand the easiest way to provide thatinformation might be a navigation applike google maps where people could seeinfection risk color codeddirectly on the map as green yellow orredto indicate the crowdedness of eachstore so by this timewe had this tantalizing idea and apossible way to make it workbut we were also mindful of thehippocratic oathfirst do no harm because we know fromexperience that even with the best ofintentions things don’t always work outthe way we expect them tothere can be unintended side effects andfeedback loops andthese things only show up when a systemis deployed into the real worldso before unleashing crowd meter we wantto understand if there are circumstancesin which it will cause more harm thangood or in what situationsum it could be most effective and thenuse those findings to inform ourspecificdesign to ensure crowd meters is builtand usedin only beneficial ways so for exampleif only a fraction of people are usingthe appit could really help them but ifeveryone uses the appit could potentially backfire so if theyall rush to the safeststore then it suddenly becomes the mostdangerous store and everyone there getsinfectedand to make things even trickier thissituation could arise due toa media feedback loop so if the mediareports on rising infection rates thenwe couldsuddenly see an increase in crowd meteradoption uh possibly creating this kindofoverload situation so how do we figureout how todo no harm without putting people atrisk in the processso as a start we run simulations whichare basically computerized mock-ups ofthe real worldthat capture the important parts of whatwe want to understandthere are many covid simulators alreadyout therebut our simulations are more complex andhave unique requirementsbecause we’re trying to understand whathappens when you introducean online app that impacts how many ofthese different systems interact witheach otherum so umso we have a model of humandecision-making behavior that balancesrisk aversionagainst the burden of traveling farthera model of the virusa model of disease spread populationflow model of human movementsand a model of how crowd meter uh wouldworkum but simulations are only as good astheir underlying assumptionsuh garbage in garbage out so we gathereda team of specialists to help ensurethat these fiveinteracting models approximate realworld conditionswith kind support from aspen tech policyhuband a donation of azure cloud computingtime for microsoftwe built a crowd meter simulator to helpus test and refine the crowd meterconceptbefore we go live and this is what thesimulator looks like so this gridrepresents a hypothetical townand then the three big dotsare grocery stores where the color ofthe store tells how crowded it isand the tiny dots are people so greenpeoplerepresent healthy and red people areinfectedand then any dot with a little blackoutline around it is a simulated personthat’s using the crowd meter appsince our main goal is to flatten thecurve the first question we ask thesimulator is whether or not crowd metercan actually do thatso flattening the curve doesn’tnecessarily mean reducing the overallnumber of people who eventually getinfected by covidit means to slow down the infection rateso that our health care system does notget overloadedand people who get coveted can actuallyhave better outcomesand until we have a vaccine this is ourbest option and even with vaccinesit can still help so we ran oursimulatorwith a range of crowd meter adoptionlevels and the first thing we noticedis that with more people using crowdmeter we were actually seeing a moreeven distribution of people going todifferent storeswhich is exactly what we’d hoped fornext we looked to see if the curve gotflatter when people were more evenlyspread out over storesand we were quite surprised to discoverthat it didn’tbut we soon realized that our initialassumption of 10 000 people sharing onlythree grocery stores in our simulatedtownresulted in a very high person densityat all three stores all the timethis meant that even when people went tothe safest store they were almostcertain to become infectedso in that circumstance using crowdmeter to distribute people evenly acrossstores gave no advantagesbut when we adjusted the population sizein our simulator to real world valuesabout a thousand people per grocerystorewe saw exactly the kind of curveflattening we expecteduh so crowd meter was working and thenand and so then the next question wasum if it can work is there such a thingas too much crowd meterso we ran more simulations at very highadoption ratesand we discovered that infection ratesdo start to go back upwhich made us realize there could besuch a thing as a crowd meter overdosewhere adoption levels could get so highthat we start to see an oppositeor what they call a paradoxical effectso we thought the reason this happens isdue to the travel delaybetween when people decide to go to astore and when they finally arriveby which time the store could be morecrowded so this made us wonder how datalatencywhich is the freshness of the crowdmeter information can impact how wellcrowd meter worksand this is important because thepopulation flow data sources availableto usfor the crowd meter app are notperfectly real time there are somedelaysso we decided to test this explicitlyagain by looking at different levels ofdata freshnessto see if crowd meter could still behelpful even if the data were notperfectly real timebut instead near real time for examplewhat happens if the crowd meter data is15 minutes old or 30 minutes oldhow does that impact the flattening ofthe curve and we discovered that withmoderate data delays crowd meter couldstill be helpfulbut beyond about 40 minutes of delay itcould actually increase infection ratesso observing in our simulations that thecombination of high adoption levels anddata delays could actually make thingsworsewe set out to examine the full space ofthose conditions to get a better senseof where crowd meter would likely behelpful or harmfuland this set of simulations revealed asafety zonedepicted in green here and a danger zonein redfor operating crowd meternow all the simulations so far werebased on covid19but we hope that crowdmeter could beused in the future to help attenuate thespread of other viruses such asinfluenzabut each kind of virus has its owncharacteristicsso we set out to also simulate theimpacts of crowd meter on viruses withdifferent levels of infectivityand discovered that for covetedinfluenzait could be most effective um andpossibly for the common coldbut for highly infectious diseaseslike the measles even crowd meter can’tcontain the spread and situations likethat will require strict quarantinesso what do all these findings mean wellwe learned that under certain extremeconditionscrowd meter would not help at all forexample when there weretoo many people uh in all the grocerystores when the hot when there was aan unusually high ratio of people togrocery storesuh which could happen you know if halfthe grocer shut down suddenlyor when the virus is highly infectioussuch as with the measlesbut the good news we discovered is thatfor the current virus covet 19and the typical conditions found in mostplaces crowd meter will flatten thecurvewe also learn that crowd meter like anyintervention has a therapeutic rangefor example antidepressants can actuallyworsen depression at the wrong doseso similarly if we if we think of crowdmeter like an anti-pandemic pillwe’ve observed a similar paradoxicaleffect where increased adoption isbeneficialup to a certain point after which it canactually cause harm when data delays arepresentso what’s next so that green area in thefinal graphrepresents crowd meters therapeuticrange and so our next step is to releasea pilot version of this app in a contextthat falls within the safe green area ofthat graphwith participant excuse me withparticipants who voluntarilyand knowingly agree to contact tracingso we can test our assumptions andsimulation outcomesagainst a real world implementation andthen we’ll use those results tofine-tune crowd meter before releasingit into the wildwe also intend to run new simulationsthe test of crowd meter could be evenmore effectiveby evenly distributing store visits notonly between storesbut also at the same store overdifferent times of the day and then useartificial intelligence to predictpopulation flowto make that work even betterso we’re encouraged by these resultsand with crowd meter safely deployedmaybewe can more easily create a new normalthat we can all live withthanks back to you mywonderful thanks so much pedro i amthrilled uh about your project and it’sreally inspiring to see how we can buildtools to help all of us make smallbehavior changesuh and and do our part to reduce virustransmissionum so thank you for all of your yourincredible workum i’d love next like to introduce ourkeynote speaker for the eventum rachel coldy cut rachel is atechnology expert and strategist whospecializesin the social impact of new and emergingtechnologyshe is the former ceo of dot everyone auk-based responsible technology thinktankand a ceo she led influential andgroundbreaking research into howtechnology is changing societyand helped develop practical tools forresponsible innovationprior to that she spent almost 20 yearsworking at the cutting edge of newtechnology for companiesincluding bbc and microsoft i’mextremely excited to have rachel sharesome thoughts on what’s been presentedthis morning over to you rachelhi uh well i mean it’s beenamazing to see these twoprojects and think about them in thecontext of some of the work that i havedoneover the last six months umand i think the two bits of context thati canbring to that is both the work i’ve donearoundum public trust to do with techum with um technology at thisuh the time andsome um research tounderstand kind of how communities areyouare using technology and how that ischangingbehavior on the ground and i think thefirst thingi would say umfrom my perspective here in londonis that we’re in a really extraordinaryumrelating to data so data is normallysomething that moves around fairlyin in visibly people don’t talk about ita lotum and in the last six months a wholeload ofideas and con concepts about uhthey have become the things thateverybody is talking about all thetime we’ve heard just now about uhflattening the curvethere are people who have never reallylooked at agraph in their life who are now talkingabout umlog their scales and exponential growthwe’re all interested in the earth rateand it’s this kind of weird world inwhichwe’re all suddenly fascinated byan expert in things we don’t knowanything aboutin a context that we don’t understandand i think both of these projects thatwe’veseen are really interesting in how theytake the umthe certainty and the multiple and thecertainties that everybody is livingwithat the moment and find hooks for us tohangparticular activities onbecause i think as well one of thethings that’s come up inresearch that i have been um doingis that although this is anextraordinary time in thatyou know this this terrible virus isuh affecting everyoneall over the world many of us are athome alonemany of us are alienated fromcommunities where we would often findum knowledge and get contextuh and yet there’s information that isavailable to us all the timeand the role of of trust in thisi think is is really importantand so to go back to tommy’spresentationthinking about how does anybody knowwhere any informationhas come from umhere in the uk where i think we have avery i mean i think we have a verydifferentrelationship to the stateuntil maybe the start ofthe pandemici think we saw like a relativelybenevolent relationship betweenindividuals and theearth state as a whole with quite highlevels of trustquite high levels of the trust inthe information that was given outand that actually one of the things thathas happened herein the way that technology has been usedis that rather than speaking to thekinds of openness and transparency toseewe’ve heard about uh with these twoprojectsrather than kind of using the best oftechnology which is really aboutopenness andinteroperability and democracythere’s been a tendency in agovernments i think all over the worldum but particularly in the ukto think of data as a the black box tonotexplain it to not empower peopleto spit figures out and to changethe rules and i think the things thatare really interestingabout these two tools that we’veseen today is that in different waysthey give us more autonomyas individuals and communitiesone of the really one of the reallycomplicated things that has happenedin this period is theidea that having beenindividualized over decades and decadesbyyou know markets and capitalism andhousing and all thosethings and then again by our uhthe phones that actually we’re beingasked to acttogether but potentially we don’t knowhow todo that and i think at all like crabmeter is really interesting because ititit kind of helps helps us all umkind of lean on reliableexternal information i knowat the moment people find it verydifficult to work outwhat to do nexthow to um navigate advice that ischangingall the the time and i think the mixtureofbeing able to visualize and being ableto getlocal information is probably reallyhelpful thereand i’ll i’ll just speak quickly toone uh the findingum that came out of some work that i’vebeendoing over the last couple of monthslooking at how communities areum using knowledgeand one of the things that came out alot is that uhthe mutual aid um is bothgreat and it has it on the side andso the thing that we saw is where thereis notan existing communitywhere people are not already umtalking to each other where there is nooneto host where there’s no one to to kindofbuild the capital in the groupwhen um the mutual aidhas appeared in acommunity mediated only throughtechnologyit’s been much harder for people to askfor helpbecause they haven’t had those bondsthattie them together with the people aroundthemyou know it’s easy tooffer help on an app but it’s muchharder to askfor it and i think that you know thething that makes mefeel quite hopeful aboutum both of these projects is thatthey’re not trying tochange how many people behave at oncethey’re not trying to sort of bite offyou know wholeum system but theythey kind of speak really well tothe kind of technology that can bedeployed and usedresponsibly small partsthat are[Music]adaptable connective can beloosely joined withother things in in umbritain i think as i was justsaying earlier um we’re a much smallercountryyou know and there’s and i thinkthere’s an idea that it is possibleto create an enormous entrancesystem that will resolveall the technology problems uh and allthe peopleproblems but actuallyum going to an idea of small partsloosely join where people andcommunities can take the things thatworkfor them where they’re monitored andmeasured andadapted over time feels like a reallypositive route ahead as we umnavigate whether it’s the yearor two of the coming uncertaintyand i think those kinds of tools willhelp us all behavein ways that can help each other aswell as ourselves um thank youover to maya wonderful thank you so muchrachel for thoseremarks um i agree that it’s importantto think about the context we’re in andhow it relates to our projectsum and in a time where public trust andinformationseems to be at all time lows everywherein the world and especially herehere in the states building tools thatimprove trust and give information tothe peopleto inspire behavior change in some waymight be an incredible policy loverum so thank you again um another thingthat i feel inspired about by today’sprojects is that they give practicalsolutionsto the issue of providing moreinformation and improvingpublic trust which is incrediblydifficult um so we’ve tried to bringabout some practical solutions today toimprove information sharingum but if there’s anything that we lovethe audience to know here at the hub isthat there’s plenty more to doand that there’s plenty more that we cando as as engaged citizensum so now i’d like to bring back thepanelistsback up to answer questions againanother reminderthat you can submit questions in the q abox belowum so the first question um is forpietro um and this comes from theaudienceso crowd meter is a fascinating projectuh but this problem of sort of manypeople rushing to one location must be aproblem that google has looked intoin depth with things like google mapsand routing softwareum are there any insights here that aretransposableyeah that’s a really apt question ithink um soit’s it’s a really fun problem space tokind of geek out onand we were definitely mindful ofof you know techniques like reroutingeither implicitly or explicitly soyou know people will will willpotentially look at the app as they’regoing to the store to see if thesituation’s changing but we still havethisthis potential data delay problem and soi thinkum we’re certainly going to going tolook intoyou know google’s findings on this uhand and and probably from other sourcesas wellbut i think a couple uh rel relevant uhapproaches here one is that oursimulatoris a computational uh model umso we actually represent each personmoving in this townand going to stores and when we slowdown the simulator you actually seepeople driving to the stores andit really helps to sort of develop anintuition abouthow behavior changes under certainassumptions when you tweak the modeland so this is a computational model butthen once we get those insights and weget resultswe can then transform that into the moretypical form of these models which isis differential equations which is muchmore efficientand what most coveted simulators usetodayso and i think part of it is to appealto complexity science which helps usidentifykind of how different equilibrium statesarise andand then when we tweak the model youknow can we achievethe target equilibrium state where youdon’t get that rushing in effectand then i think finally um we’re we’regoing touh explore predictive modeling so theidea isyou know can we anticipate when weprovide this information on crowd meterat these level of delays um you know howare people going to make decisions aboutwhere to go and if we know thatum you know then how does that maybeshiftour recommendations you know orsomething like that but it’s very much achicken and egg problem which is whatcomplexity science is perfect forso back to you maya wonderful thankspietroum the next question is for the firsttraffic team umso what might be other sources of supplyother than fact-checking organizationswhichinherently have limited bandwidth andthis comes from spencer yen from the mitmedia labum i can take this pleasure let me knowif you want to jump inum well this is precisely the kind ofquestionthat um we we can ask through this thisframework and this approach so i’mreally pleased tohear you uh thinking about that alreadyum and i think yeah it’s an openquestion within a way it’s for all of us to thinkabout how we might answerum the ones that we’ve been thinkingabout other than fat checks becauseyou’re completely right uh there areconstraints on resourcesand also fat checks are only aparticular kind of information comingfrom a particular kind of organizationso other options are things likeresearch papers or pre-prints we saw alot of information comingthere in the in the process of thepandemic this yearwe could also look to news reports aswell umthe if we look at credible news sourcesoften these are very early to strikewithcredible information um and is is thekind of information that reaches peopleas well becausewhen we think about supply it’s not justa question of whether the information iscreatedbut it’s actually got to reach peopleand that’s where the news becomes sosignificant because they tend to have amuch biggerreach than um than fact checkersand obviously then you you you get intoquestions of tvyou know there’s our people gettingpeople uminformation getting carried to people ontv that might not be so easy for us toscrape onlinebut the news is definitely a big sourceum government statementstend to be fairly well researched that’sanother place that people might look forinformation as well and it might be thekind of information that flows intoother sources like news and fat checksand so on so that is anotherplace for us to look uh civil society aswelluh think tanks um it might be the samecharitiesare providing good information we’ve gotum an organization called umcitizens advice in the uk that’s verytransparent with its dataand also provides really incredibleinformation that’s another personanother group to look tomedical bodies the who they were doing alot of um they’re really trying hard nowi think they’ve even gota new job role called epidemiologistsuh who are designed to to look at thiskind of informationum and they’re providing a lot of thethe kind of information that’s sourcedso that’s just a kind of a quickoverview ofpossible options but i think what what ifindum most productive and i supposeexciting about this route is thatit helps create this framework to startasking those kinds of questions like howwould we gather all this information tohave a pictureof what people are um are accessing whatcredible information people areaccessingso yeah if you have any other ideas tellme because i will definitelysteal them and implement them i don’tknow whether you have any other thoughtsbut jayno i think it’s pretty much this justreflect that there are projects that trytoboth automatically and manuallyattribute bias or credibility scores tonews sourcesbut again this is extremely problematicright becauseit’s very difficult to list all theavailablenew sources and then performobjective evaluation of their bias andtheir credibility soagain it’s it’s it’s an open questionwonderful thanks so much uh tommy andpedroum pietro we have another question foryou um so a crowd meter app would likelydraw a lot of excited potential usershow would you control adoption levels sothat the app can stay effective in thegreen safe zoneyeah yeah great question um yeah itreminds me of thisthis commercial during the dot-com boomwhere everyone’s sitting aroundtheir new website goes live and they’reexcited and back then we had these sitevisit countersand it starts to go up and they getexcited and then it starts to go upfasterand then they start to look at eachother and pretty soon it’s spinning outof controlum so um i thinkyou know our our idea is that umif we can implement um predictionin the model so that umwe can mitigate the effects of delayeddatathen the adoption rate issues umgo away so in other words it you know ina perfect world with instantaneous dataor perfect knowledge about what peoplewere going to do where they were goingto gocrowd meter would work great with 100adoptionum so obviously we don’t live in thatperfect world but i thinkyou know from our model of crowd metertoday to where we think we can go i meanthis is just the veryearliest you know sort of mostrudimentary implementationum and i think the sweet spot for thisone is 20 to 40 adoption well if we canget 20 to 40of the people you know in the world inthe us and any country to use thisthat’s a huge adoption level for any appand and i’m i’m i’moptimistic that with the experts i’mworking with and the methods we’re usingthat we could support a much wider greenzone that encompasses a much higheradoption rate as we implement these newmethods and that’s what we’re going tobe testing out in simulationas wellthank you wonderful thanks so much petroanother question for the first strapteam um so your projectis identifying data deficits but whatcomes after thisthat is how do you think this projectwill inform first drafts othermisinformation fighting methods and whataresolutions that come out of thiswell um that that is a great question imean with this wasthis was born in the context of thepandemicwe are not so much in that immediatepost-crisis phase anymoreum so there’s a there’s a couple ofthings for us to think about when wewhen we think about taking it next firstof allhow can we be best prepared for the nextcrisisum i was reading recently that there’sprediction that will havea similar kind of health crisis maybenot the same scale but something like itevery five or so years so i think one ofthe questions is what can we have set upin placeso that we um are better prepared forthe next crisisthe second route is well outside thecontext of crisis hack we make thisusefuland i think the answer to that is um bylooking at specificsubject specific topics so the key thingthat we’re looking at the moment isvaccines obviously that’s the nextphase of this epidemic as it’s beencalledso what we want to do is is break downvaccine conversationsinto sub-topics so that we can trackthem moreprecisely and um start to use thesekinds of metrics that we’ve talked abouttodayto understand what information peopleneed and then connectwith um information providersto help fill that gap um so it might beum it could be uh umback checkers obviously it could be umsocial media platforms it could becommunication agencies working withgovernment bodies like the ununicef um it could beinfluencers i mean they’re a huge factorof misinformation at the moment so wecould be trying to helpum influencers on instagram anddifferent social media platforms likethatum answer the questions that their umtheir followers are likely to have andbe a credible information source ofthose peopleum so i i think yeah we what we’re mostlikely to be looking towards is how wecan break this down for different topicsum monitor these conversations closelyand then connect with the right peopleto provideto fill those gaps to kind of addressthose deficitswonderful thank you tommy um rachel ihave a question for youum so you spoke a lot about theimportance of giving the publicinformation and these projects are aboutcrowdsourcing informationdo you have any insights on how we canuse tech to facilitateuh crowdsourcing information withoutsome of its negative consequenceslike misinformation popping upi mean i think uhgosh it’s complicated i meanuh just to go back to what tommy wassaying i think one of the reallyimportant things here is to look attopicsuh in the contextum and i thinkat the moment it may be less aboutcrowd it i think it may be more aboutbuilding the ability to trust and toquestionand i think almost making a culture inwhichit’s okay to say i don’t know i don’tunderstandi need help um becausei think potentially one of the dancesides of ofall this information again is that itfeelslike we all to know umyou know and that’s what i i was sayingpreviouslyabout how everybody’s now anan expert in uh data oran epidemiologist because we’veread about it on the internet like ithinkactually where i would go is make itmuch more about the criticalskills and the confidence to say umyou don’t know rather than making itbe about collecting more and moreinformationif that makes sense thank youyes that definitely makes sense thanksso much rachelum so we always like to ask one lastquestion here at the hubso if you had a one-line bumper stickerwhat is the main takeaway of yourprojectin one line um so i’ll go in order ofour panelists of how they show up on myscreenum so pietro what is the one linetakeaway of of today andand your project okay umso i i mean i think i think nobody wantsto get sick from covidbut we all uh i think value ourautonomy and transparency of informationand we now have evidenceum that crowd meter can help reduce theimpacts of kovidby keeping us all informed of ouroptionsso we can achieve a new normal that wecan hopefully all live withwonderful thanks pietro uh tommy what isyour one line takeawaymy takeaway would be the informationis a public utility we depend on it ourlives depend on good informationand as a result we need to find ways tomonitor itjust like we monitor other things thatwe we rely onum you could compare it to somethinglike water orthe weather or money like we need tomight be able to monitor these thingsand this the thinking in terms of demandand supplies is likeone entry point into trying to do thatgreat thank you uh pedroi would say that um i mean the territorythat we’re navigating hereis still very unknown andwe are still in the first steps ofunderstanding howinformation is relating with uh with aglobal health crisiswonderful thank you and racheluh i i have spent a lot of the lastsix months kind of looking at the waysthat technologies aren’t helpinguh looking at you know the problems ofthe surveillance and lots ofuh the privacy and i think the thing iwould take awayfrom today is that it’s like reallyhopefulthat actually um it’s reallygreat to see practicalrealizable things that are being thoughtaboutum responsibly rather than in it youknow notall technology has to be about totaluh the surveillance and so that’sum really optimistic thank you for thatwonderful uh thanks for ending on on theoptimistic note i always appreciate thatum so i just wanted to before we end iwanted to thank umfirst the funders that made these covidchallenge grants possibleso that includes the ford foundation andthe omidyar networkum i wanted to thank the aspen digitalteam here at the aspen instituteuh that have been working behind thescenes so that includes our directorbetsy cooper uh mayhawesevilla carner and javier i of coursewant to thank the grantees themselvestommy pedro and pietro um i want tothankour keynote speaker rachel and i alsowanted to remind you all that ournext and last webinar of the series willbe on wednesday october 21stat 9 a.m pacific time or 12 p.m easterntimeum so we’ll be presenting on our lasttwo grantee projectsfocused on informing covid response andcommunications through techum so we’ll be hearing from michaelmorrissey from muckrock who will bepresenting a product that helps quicklyscaleefforts to collect and analyze dataneeded to inform policyand community responses to the covidcrisis and we’ll also be hearing frombrendan lind of human agency who will bepresentingon a responsive messaging service thathelps connect individualsto specific resources in theircommunities so i hope you all will joinusand in the meantime we wish you the bestin these difficult timesthanks so much for joiningyou
The COVID-19 crisis has posed myriad risks from all directions–from the spread of a dangerous virus to the spread of threatening viral misinformation online, understanding how to navigate a pandemic on- and off-line is a challenge. How can we build tools to aid informational understanding and response in pandemic conditions?
Join Aspen Tech Policy Hub Challenge Grantees as they showcase their projects focused on the theme: Mapping On- & Off-line Risks in a Pandemic. Following the presentations of the projects, technologist Rachel Coldicutt gave further remarks.
Presented Projects
COVID-19 Data Deficits Dashboard by Tommy Shane and Pedro Noel: First Draft has built a tool to make the world’s credible coverage of coronavirus, such as original reporting and debunks, more useful to journalists and researchers. They have leveraged data science to visualize where existing credible information is flowing globally and where it is not, so that reporters can identify information needs.
The CrowdMeter Project by Pietro Michelucci and the Human Computation Institute: CrowdMeter is an app being developed by the Human Computation Institute to help people reduce virus transmission by making informed decisions about where and when to run their errands; for example, someone who wants to buy groceries might be willing to drive an extra mile if that seems like a safer option. The goal is to help achieve a sustainable “new normal” by aligning individual goals with population outcomes.
THE BEST OF ASPEN DIGITAL, RIGHT IN YOUR INBOX.
Sign up for our email list to stay current on our work.