© CelPress Cell Leading Edge Infodemics: A new challenge for public health ‘Sylvie C. Briand,’ Matteo Cinelli,’ Tim Nguyen,’ Rosamund Lewis, * Dimitri Prybyisk,” Carlo M. Valensiso,° Vittoria Colizza,’ Alberto Eugenio Tozzi,” Nicola Perra,’ Andrea Baroncheli, Michele Tizzoni,"” Fabiana Zollo,” ‘Antonio Scala,'='° Tina Pumat,° Christine Gzeriak," Adam J. Kucharski,'“ Akhona Tshangela,'° Lei Zhou, ‘and Walter Quattrociocchi'™" ‘Global Infectious Hazards Preparedness Department, World Health Organization, Geneva, Swtzorland Departmen of Environmental Siences, Informatics and Statistic, Ca! Foscari University of Vice, 20172 Venice, Hay ‘impact Events Preparedness Unt, Global Infestous Hazards Preparedness Dopatmert, ors Heath Organization, Gonova, Switzerland “infodemic Management Group. Health Emergencies Programme, World Health Organization, Gnova, Switzeriand Global immunization Division, Genter or Global Health, US. Centers for Disease Control and Prevention, Atlanta, GA 90080, USA ‘Errco Fermi Research Contr, Pazza dl Viminalo,t ~ 00184, Roma, tly "INSERM, Sorbonne Urivesits, institut Pore Lous Epidémiologie ot de Santé Publique, IPLESP, Pars, Franco Mtutitactoral and Complex Diseases research Area, Bambro Gast Children's Hospital, ome, aly ‘Networks and Urban Systems Contre, University of Greenwich, London, UK "Department of Mathematics, City University of London & The Alan Turing nett "si Foundation, 10126, Tur, aly “Appice Lab, CNR-SC, Rema, aly *5ig Data in Heath Society, Roma aly ““Dopartnant of Infectious Disease Epidemiology, London School of Hygione& Tropical Medicine, London, UK ‘sAirica Centers for Disease Control and Prevention, Aiican Union Headquarters, Adis Ababa, Ethiopia ‘Publ Health Emergency Certer, Chinese Center for Disease Contr! and Prevertion, Baling, China "Department of Computer Science, Sapienza Univers of Rome, 00185 Rome, aly *Corresporcence: water cuntwocecchi@uniromat i btpss/ldterg/10.1016/,co202.10.031 London, UK. ‘The COVID-19 information epidemic, or “infodemic,” demonstrates how unlimited access to information may confuse and influence behaviors during a health emergency. However, the study of infodemics is relatively new, and little is known about their relationship with epidemics management. Here, we discuss unresolved issues and propose research directions to enhance preparedness for future health crises. Information in a disintermediated access to information may influnce non, which seems to be strongly related ‘environment {events unfolding in the physical world. to the evolving business model of infor- The COronaVis Disease of 2019 Our grasp of what an “Infodemic” Is mation clsseminaton, currenty. domi- (COVID-19) pandemichas shown the t= and how it happens is sill shallow and nated by social media (Cinali et i ical role of information difusion in heath evolving. A possible reason forthe mist 2020). In ight of ths, “infodemic” was ‘emergencies. and ofsis_ management. around the term “infodemic” may reside defined as “an overabundance of infor- ‘The complexities of science knowledge ints very nature ofan inutve umbrella mation—some accurate and some not— ‘and the scientific method are dffcult to term that, however, includes many ramil- that occurs during an epidemic” (T2na- ‘communicate to a broad audience— cations ranging from communication to charoensathien eta, 2020). ‘especially in ight of heterogenety in sci- epidemiology and that Inks to several The uptake of an infodemic definition ‘ence literacy (Fuhs, 2019). This context open scientific debates such as that on that does not cirectly tackle the concept tends to yield extreme oversimplficatons misinformation spreading and its effects of misinformation seems to point toward (@9,, 2210 rskierteal risk) that may on society (Sinn and Camargo, 202"), a bigger problem than the “true vs flea” polarize natives. At the same time, the To be more specific, the term “info- dichotomy, natualy limited by the fact ‘ongoing revisions of evidence and an demic," Intended to mean an “epidemic thatthe concept of truth may fer across ‘overabundant and changing information of information,” was introduced by "ic social groups. For this reason, other landscape may induce confusion for pol- Kop! (2003) to define the amplification aspects related to human behavior, leyrmakers and civi society (Galotti effect of the news about severe acute particulary the tendency of individuals to {8 a, 2020; Tangcharoensathien et sl, respiratory syndrome (SARS) due toinfor- select information confirming ther beliefs 2020) For example, the problem of vac- mation technologies. A more recent def and ignore asserting information (De! Vi- cine hesitancy seems to be elated to an nition of “infodemic" emphasized the cariostal, 20°), shouldbe taken ntoac- ‘roded trust in institutions fueled by element of misinformation spreading count to better understand the infodemic rmsinformation spreading. Such a pro- rapidly through social media platforms process, Indesd, itis reasonable to hy- ‘cess has been called an information and other outlets (Zarocosias, 2020). pothesize that behaviors are important ‘epidemic or “infodemic,” and itis a clear However, misinformation does not fuly for infodemic development as much as ‘example of how virtually unimited online capture the complexity ofthe phenome- they Impact the epidemic sphere by 6010 Col 124, Decombor 8, 2021 © 2021 Eleovior neCell atering the chains of disease transmis- sion (6, using masks, teleworking, and ‘quarantning to reduce infection spread) However, itremains tobe fully established the extent to which the infodemic affects behaviors thet, in turn, may be relevant to the dynamics of ‘pandemics. The apparent analogies between epidemics ‘andinfodemicshaveledto the suggestion that scientists and policy-makers can investigate, model, and monitor the two phenomena similar (Scales etal, 2027) Here, we argue that an infodemic, though intertwined with an epidemic, i {2 distinct phenomenon resulting trom many interacting and overiapping pro- ‘cesses such asthe production, consump- tion, and amplification of (potentaly harmful) information ealine. Thus, these Processes’ analogies, differences, ond interplay must be considered to develop practical guidelines for managing and preventing future epidemics. This paper ‘alms to spark dscussions and collabora- tion to enhance preparedness for future health crises and improve the ability to anticipate the economic and. social impact of pocies put in place, We start by outlining analogies between epidemics ‘and nfodemics, as well as eitcal ifer- fences. Furthermore, we highlight the Importance of considering them as mutu= aly dependent and interacting phenom- ‘ena also pointing out specie challenges for the future, Hopeful, a deeper under- ‘standing ofthe relationship between info- \demics and epidemic willed to insights that alow early prediction of epidemiolog- ical trends and effective management of ‘communication during epidemic out- breaks and vaccination campaigns. ‘Analogies and differences Information dfusion and social contagion processes are often characterized using {epidemic or epidemic-inspred models, to the extent that rapidly sharing informa- tion is said to be "going viral.” However, information and epidemic spreading also ‘ental citcal diferences. Agent Epidemic. For epidemic processes of in- fectious diseases, the root cause is iden- tifable ina pathogen with certain bilog- ical features. (e.9,, infectious period, transmissibilty, asymptomatic forms of infection) that spreads in a population through contacts between humans. Dur- ing an epidemic, infectious agents belong toa single strain, athough variants ofthe ‘original agent may emerge over te. Infodemc. For what concerns an info demic, the agent's represented by ames ‘sage broadeast in given medium (e.g. Particular conspiracy theory in a vi ‘Video, a statistic on vaccine efficacy in a tweed). Such a message can appear in ‘many forms depending on the communi- cation channel and not necessarily be deceptive or harmful. Unike the relation ship between a pathogen and a host, the message Is subject to the audience's interpretation and it can be perceived as hostile due to cognitive processes, misin- terpretations, andthe way it is presented fon the medi, Medium Epidemic. For epidemics, the medium is represented by routes of transmission (eg, respiratory, ora-fecal route, sexual Contacts) whose involvement depends ‘also on the pathogen which mainly drives prevention strategies. Infodemic. In the case of infodemics, the medium is te communication chan- rel. Conversely from the case of epi- dlomics, the possible set of media able to fv! infodemic processes is constantly ‘evolving with agents (messages) possibly navigating and mutating from medium to medium. Furthermore, whie some media are subject to moderation that reguate their functioning some others are nearly Impossible to montor (e., dark web ‘communities. or closed online groups), thus making some routes of transmission essentialy unobservable. Another subtle difference between epidemic and info- demicis the possibilty ofmedia platforms totune the amount ot information that one ccan retrieve both about them and about the agent under investigation. Two rele- vant examples are the case of Facebook narrowing the possiblity to collect user data after the Cambridge Analytica issue or the case of Gab, an independent social media platform, accusing academics to ‘smear thir platform with a consequent downgrade of their API. Furthermore, due to phenomena such as media fragmentation, the user base can be ‘extremely heterogeneous across comms- ration channels as diferent media sone ifferent purposes (exchanging opinions, © CelPress ‘watching the news, gaming) and entail a ferent lavel of attention to gather rela- vant information (watching TV versus reading a news article). Such a strong fragmentation of hosts forthe pathogen ‘acs transmission routes seems to be more nuanced in the case of epidemics. Finally, wile the number of routes of transmission fs essentialy open ended for the infodemic, itis not the case for ep- Idemics (routes of transmission. are nether infiite nor continuously evolving withthe pathogen) ‘Timescale Epidemic. The timescale depends on ‘several factors, such as hurnan behaviors relevant to the pathogen's transmission route, the pathogen's biological features, and the immune response of individuals. ‘The evolution of an epidemic is usually made up of subsequent waves due to the accumulation of susceptible incvid- uals over time. For example, according to simple Susceptible nfected-Recov- ‘ered-lke models, epidemic waves take ‘off when the basic reproductive number ‘ofthe dlsease—a function of the features (ofthe virus, the contact patterns, and the fraction ofthe population that is susceptl- ble—surpasses the etical threshold. The infection of individuals, then transitioning to an immune or recovered state, leads ‘overtime to an epidemic wave, followed by a decreasing incidence of cases given acimiishing proportion of susceptibiein- ‘viduals. Beyond the natural evolution ot ‘an epidemic, waves may occur when population immunity is partial and pull health measures are temporary. Upon ‘the cessation of non-pharmaceutcal pubic health Interventions and. social measures (Pe=, 2021), susceptible pop- Uiation members are more tkely to be ‘exposed to the pathogen (> Domenico etal, 2020, Infodemic. At the moment of wring, the timescale of an infodemic is a key factor that stil requires a posthoc invest- ‘gation inorder to be formalized. Nonethe- less, pevious stucies displayed how the ‘dynamics of information spreading are based on social contagion, the spread of ideas, atitudes, norms, or behavioral pat- ‘tems from person to person trough s0- ial influence, intaton, and conformity. ‘Social contagion depends on users att- tudes, tendencies, intentionality, social cat 184, Docomber9, 2021 6ott| @ CelPress influence, and ties; therefore, its strength ‘and duration are likely to depend upon ‘several factors, such as pre-existing be- liefs, polaraation on a topic and its ‘complexity, the extent to which informa- tion is considered reliable, and users’ ‘engagement. Considering such dy- namics, the wavy evoktion of an ‘epidemic could be ciscogarded by the in- fodemic, whereas the production of new information regarding the disease could be continuous and potentially decoupled ‘rom epidemic waves, Network of Interaction Epidemic. The process of epidemic ‘spreading rales on the patterns of con- tacts along which transmission occurs in- vials who have a connection are at risk for transmission). For epidemic pro- ‘cesses, physical proximity and/or interac- tion with the agent or nfectious hosts are requiredto transmit the disease by means ‘of specific behaviors such as social gath- ‘ering, sex, etc. Networks of interactions land average daily contacts may be het~ ferogeneous across geographical areas {the higher the number of contacts in a time unt, the higher the velocity to develop and spread the infection), and non-pharmaceutial_ interventions | can ‘substantially reduce the density and tre- quency at which (new) links occur Infodemic. n the case ofan infodemic, the transmission is remote andhas no po: tential boundaries in trms of geograph- ical scale, Furthermore, information ‘spreading is becoming more and more rapid due to dsintermediated communi- ccation and content production allowed by social media. Therefore networks of interaction, such as those underiying so- cial media, are beyond any geographical boundary and reshape continuously around new topics wih users segregating inecho chambers (here questionable in- formation prolerates with more ease) ‘and with recommendation algorithms that may reinforce shared naratves and foster individual polarization (Cire fo al, 202%), Relatedly, the current technological infrastructure for informa tion cifusion leaves room for strategie ‘campaigns attempting to influence the in- formation ecosystem (information opera- tions) where certain agents may act as (superspreaders (Ferrars eta, 2016). 6012 Coll 124, Decombor 8, 2021 Contro! measures Epidemic. Depending onthe avalabilty of 2 vaccine, immunization campaigns and ron-pharmaceutical interventions repre- sent a milestone for controling and possibly stopping cisease epidemics. Furthermore, epidemiologic surelance Can be implemented efficiently to protect the risk categories and or contain out- breaks by guiding preventive interven tions that may target specific population ‘groups. In general, according to the bio- logical characteristics of agents androute of transmission, several controlmeasures canbe implemented, suchas isolating the source of infection and preserving individ- als from contacts; protecting individuals through NPI; or implementing mass im- mmurization or targeted immunization when the population triskis clearly ien- ‘able Infodemic. The case for immunization campaigns and immunity to an infodemic is less clear. Investing in bullaing capac- ites for eritcal thinking (e.g, science Iteracy and media iteracy) and other practies such as pre-bunking (o0z=n~ eek el, 2020) may help to blunt an in- fodemic’s negative impact. However, the potential impact of an. “immunization inan infodemic, intended to encourage health-positve behaviors when facing polarization and conspiracy ‘theories, is debated (Sover et ol, 202%) National or community-wide movement restrictions, and other public health mea ‘sures such as isolation or quarantine, do rot have a clear counterpart inthe onine realm, and equivalent restrictions may rot have a similar efect. For instance removing content or banning a user trom 8 platform may be incompatible with respect to freedom of expression and makes it more aicut for pubic heath teams to identfy population concems, Furthermore, strong moderation policies ‘may cause mass migration on other plat- forms and thus result in unintended con- sequences such as further segregation land arise of ess regulated and monitored ‘tinge communities. Nonetheless, both infodemics and opi sdemics can be counteracted individual (for example through vaccination and stal-bulling) and collectively through Physical distancing and changing social rrorms). Protectve behavior against out- breaks is associated with the nation of Cell health risk, and individual behaviors and attitudes can vary due to risk perception (On the contrary, being exposed to an Infodemic may not be perceived as a risk for personal health and safety or a8 4 threat at al, despite posing @ significant health rik, ‘The apparent analogies between spi= demics and online infodemics are summarized in Tablet ‘The interplay between infodemics and epidemics While the interplay between information bout the disease and disease propaga- tion has long been identified as a key Challenge in epidemiology, the infodemic during the COVID-19 pandemic is a clear example of how the information ‘ecosystem and narratives may influence behaviors and pubic health outcomes. How the people perceive an epidemic ‘may potentially impact choices that ind- viduals, communities, and authors ‘make for infection prevention and control, with possible adverse effects such as: ‘© Authorities’ delays in_identiving, veloping, and implementing effective and appropriate policies based on the best avalable infor- mation at the time. Ths is typical of “new” epidemics, where spread mechanisms can be very diferent from those of previously known outbreaks. Conversely, adopting appropriate policies leading to a substantial reduction of the infec tion rate may make people underes- timate the seriousness ofan issue. Early success might also lead peo- ple to cisbetieve there was an actual Issue or emergency to be ad- ‘essed. Often logical cause-effect links ar lost. © A community may reject policies ‘and expert advice. This 's more likely when an infodemic. erodes ‘rust in institutions, generally exac- exbated by economic crises and rumors or misinformation. Accep- tance of policies depends on a myriad of factors: identity, values, ‘education, and more. Proper pol- Icies may eventually be perceived 188 inappropriate due to the info ‘demic process making the outcome of polcy efforts, in the context of aCell ‘Table 1. The main analogies between epidemi Features Endemic ‘agent Infectious ager ie, vrs, bactrim, fangs, paras.) Medium Feuta of transmission (respratory,raecal rout, sonst contact.) Timescale Infectious period, reproductive numbor Network of Theater of contacts Iteraction song whicnarsmsion 004 reviuals whe hve a contact ik fortransmitiee) control ‘Actor to tho opldomic ‘measures _(\acenaton campalgns ron Pharmaceutical narvontors, ‘spidomologeal survilanc) ‘dynamic society with lower res {ance and resiience to misinforma- tion dificult to predict. Beyond the intial perception of ep- demics by authorities and ctzens, which may reduce the effectiveness ofthe very first countermeasures. from preventing ‘epidemic spreading, infodemics may un- ermine vaccination campaigns. Vaccinations have always been a cen- ‘ral topic in mis~ and disinformation ‘environments, often related to conspiracy theories, and infodemics may increase the number of people befeving in such theories, especialy from the pool of individuals with low epidemics-elated culture (eg. low understanding of bilog- ical immunization mechanism, or mathe- matical effect of exponential growths), leading 10 ess diffused immunization ‘Another issues related to adverse events following immunization (AEF), whose ‘occurrence may lead to @ negative hype about vaccine safety. Therefore, since Information signals and trends may precede and foreshadow variations in vaccination coverage and incidence ‘of vaccine-preventable diseases, the behavior of the time lag between these ‘events is fold for futher study. There is significant potential to develop social listening methods for openly shared infor- mation to feed into epidemioiogical sur- vellance and early warning systems. Tnedeies Type of moseago na avn medium a, partcuar conspiracy thoory Fa wal veo, taste con vaccine efeacy inate) ‘Commuriaton channel (rows, oc media lator, newsgroup, rai program blog.) ow que information prea and accumulates, Including the persistonce of ‘isan -nfornation User and content interaction rods of ansisson,intereonneciedness, custorng, Homophiy, cortent tering lgrtins) ‘ations tol th fodemic (sk builing scence an ‘modi tracy, pre bunk] Conclusions The intersection between infodemics and epidemics represents one of the most et- ical areas for future studies to improve pre- redness and popuaton heath globally, Indes, social media racially changed the ‘mechanism by which we access informa tion and form our opinions. We need to better understand how individuals acquire (or aveid information and how those deci- sions may influence their behavior, Despite the several projects and intaives: ‘med at providing comect information to Users, the impact of this infermation on personal choices i stil an open issue Infomation consumption patterns may not necessarly be a reliable precctor of Dbehavioral_change. Instead, stronger activity on social media seems to resutin further polarization. Henceforth, incuding the complexty of human behavior in ‘epidemic management i of pivotalimpor- tance to address the many facets ofthis Phenomenon through a scienifcaly {grounded approach in order to support the design of effectve communication sirategies and develop the to's required to property manage infodemics, To reach ths goal and capture the over all dimensions. of epidemicrinfodemic ‘management, we substantially noed an Interdiscipinary approach Involving epi domiologiss, data scientist, physicists ‘and mathomatcians,rsk communication practitioners, behavioral scientists, public © CelPress health professionals, representatives of affected communities, and ideally sup- port from the leading data. providers {@.9, soctal media entities) ‘Along this path, to actuate timely re ‘sponses to cical scenarios, the scient- fic community should identity the most ‘sultable communication strategies anc provide guidelines for journalists and Stakeholders to communicate complex issues toa broader audience to avoid po- larization In turn, @ communication effort ‘should go in the dlrection of clartying ‘the roles of diferent stakeholder groups for a whole-f-society response. Even technical limitations have to be overcome with the aim of combining data and ana- lytic and social sensing o promptly iden- ty, ful, and eventually forecast social ‘ends and information voids. ‘To summarize in global pandemics, formation shapes perceptions and may influence choices and thus policy design ‘and_social response, To enforce an improved epidemics sunellance, we ‘should consider the contemporary pres- ‘ence ofinfodemics and epidemics dimen ‘sions, accounting for their singular anc ‘Shared features. W.0,,A8, and ackronedge the 100689210 Projet “Global Healy Secury Academe srowidges he Lange Proj ofthe 1t Fo ston findod by ORT Foueson, SB, RL gandaot OP. asa manta of he US Car ‘Sa member othe Chee US Gantars for Ds ‘sae Contr and Provertion Those authors lone ‘respon forthe vows exoossed 8 Be ber arty Se otrepreaon evs aor DECLARATION OF INTERESTS Cel M, Custrotea We, Galas A, Vaan 60, OM, Buprl E. Schme AL, Zola, P oto, F, and Scala, A. 2020) The COVD-8 so almedia node Se. Rep. 10,1508. Cnet, De Frenlse Moa, G, Clea A. ‘Guatroctcei, Wend Sti, Me U2), Tha ‘cto charbe eect on sot mada Pre. Na. ‘oad. Sl USA 17, 62023201118 al Veo, M, Beh A, Zo, F Paton F. ‘Sea, A. Cadel, Stwiey, HE. and (Got 184, Docomber9, 2021 6013@ CelPress ‘uateoioco W. (216. The sprecang of msn- 53-658, DF Domenica, L, Plo, G, Sabbatn, CE. ‘ost, Pv, an Cain, V (2020. Impact ot lcekdown on COWD-18epdomisin Deco France and possbl et stateges, BMC ed. 18,20, Fora, E,Varol, 0, Oa. Mensser, F and Fan A. BOT Theres of soc bes, Com Clot Val, Cataldo, N, Sec, Pan ‘Deore, Mz, Assan he rsh fh ‘osenize’ In veponse to COVID-'G epcemic. Nat Hon. Saw 1285-1265, ‘era. 2021).Non- pharmaceutical meventons ‘ang te COVD-18 panda: A review. Py. Fp 8131-82. ozone J vn derLinon.S, and Nyaren T (2020, Preburkngetewentons based on rece 6014 Call 124, Decembor 8, 2021 lata orcanrcace suscep tomietr= Ison seoseeuires Harerd Kenedy Stoo (HK) Mlnermaton Revie) s/o 0, ‘vor 2020008, Fete, Dl. 2005 When he Buzz Btee Beck, “The New You Times. Nay 11, 2008 is/ we westintorpestconvarchieopics/208/ Dahmer the arb Dackoceatabt {ote G8 0277261 Fats, 0. 2019, The msevomaton machine Se ce 36,348, ‘Scale, . Gorman. and Jamieson KH. (2020) The Covi 19 fogomi~ Appia the Epdemo= lege Mods to Coutar Mlsnloraton, N Eng ‘Simon, FM. and Camargo... 2027) Autopsy of ametaphor: The ops, use anc nd spot ot ‘ho ‘riesomic’ new media &socay, 122. ripeisor to Prva eatey ste, Cell Commentary ‘over A. Kasson L.C. Af J, Ln, Mt nd Lowancowsiy, S. (221). Uniting to ‘ngoge i bonvers that protect apart COMI 1" the roe of conspracy bells, ust and cndorsement of complrertay and alive ‘Tangoreroersatie,V. Calla N, Nowen, Pumat, T, DAgosine, M, GarieSaso, Lume, Masher A, Hamiton ABSA. [Ry ot ak (2020) Framework for menaging the (Covo-1@ hrodens: meade and eau an ‘ning rowdeouced IO tecvieaconstston. Zarosostas, (220, How oft an ifs