How social interactions mitigate extremist views

Soci­ety is becom­ing increas­ing­ly divid­ed, and we are con­tin­u­ing to see the emer­gence of extrem­ist views around the world, be it with regards to top­ics like pol­i­tics, reli­gion, or cli­mate change. While there has been a great deal of research into how this phe­nom­e­non, which is known as ‘polar­i­sa­tion’, has evolved less atten­tion has been paid to under­stand­ing how social inter­ac­tions can lead to the oppo­site effect – “depo­lar­i­sa­tion” –, which occurs when indi­vid­u­als begin to mod­i­fy their opin­ions so that they are less extreme.

To address this ques­tion, Jaume Ojer, Michele Starni­ni and Romual­do Pas­tor-Sator­ras, from the Depar­ta­ment de Físi­ca, Uni­ver­si­tat Politèc­ni­ca de Catalun­ya and the CENTAI Insti­tute in Turin, have pro­posed a new “social com­pass” mod­el to study how opin­ion varies between groups with extrem­ist posi­tions and how these opin­ions might be depo­larised¹. Their the­o­ret­i­cal frame­work has been val­i­dat­ed by exten­sive numer­i­cal sim­u­la­tions and test­ed using data from opin­ion polls col­lect­ed by the Amer­i­can Nation­al Elec­tion Studies.

“Polar­i­sa­tion may con­tribute to widen­ing the polit­i­cal divide in our soci­ety, ham­per­ing the col­lec­tive res­o­lu­tion of impor­tant soci­etal chal­lenges,” say the researchers. “It could even encour­age the spread of mis­in­for­ma­tion and con­spir­a­cy the­o­ries. Our depo­lar­i­sa­tion frame­work could pro­vide solu­tions to these soci­etal ills.”

Mod­els describ­ing polar­i­sa­tion are based on mech­a­nisms as diverse as homophi­ly, bound­ed con­fi­dence or opin­ion rejec­tion. Until now, the depo­lar­i­sa­tion process in a pop­u­la­tion has gen­er­al­ly been mod­elled for the sim­ple case of an indi­vid­u­al’s opin­ion on a sin­gle sub­ject. In real­i­ty, how­ev­er, an indi­vid­ual gen­er­al­ly has opin­ions on sev­er­al sub­jects at any giv­en time. A mul­ti­di­men­sion­al mod­el­ling frame­work is there­fore need­ed to bet­ter describe how opin­ions evolve.

When mul­ti­ple sub­jects are con­sid­ered, a num­ber of fea­tures emerge. The first is align­ment, that is, the pres­ence of a cor­re­la­tion between opin­ions with respect to dif­fer­ent sub­jects. For exam­ple, peo­ple with strong reli­gious con­vic­tions are more like­ly to oppose abor­tion leg­is­la­tion. The prob­lem with cur­rent mul­ti­di­men­sion­al mod­els is that they neglect this inter­de­pen­dence between dif­fer­ent sub­jects, which means that they fail to clear­ly describe opin­ion polarisation.

The key idea of the social com­pass mod­el is to rep­re­sent opin­ions in rela­tion to two top­ics locat­ed on oppo­site sides of a polar plane. The angle of the plane rep­re­sents an indi­vid­u­al’s ori­en­ta­tion as regards to the two top­ics, and its radius express­es the strength of the atti­tude (or ‘con­vic­tion’).

“This polar rep­re­sen­ta­tion nat­u­ral­ly allows us to for­mu­late the key hypoth­e­sis of our mod­el, name­ly that intran­si­gents with extreme opin­ions (or strong con­vic­tion) may be less like­ly to change their opin­ion than indi­vid­u­als with weak con­vic­tion,” explains Michele Starni­ni. This hypoth­e­sis is intu­itive and con­sis­tent with obser­va­tions made in exper­i­men­tal psy­chol­o­gy. “Such a polar rep­re­sen­ta­tion is very com­mon in physics, but not so much in the social sciences.”

Inspired by the Fried­kin-Johnsen mod­el², the researchers stud­ied how social influ­ence can affect the ini­tial opin­ions of indi­vid­u­als. They found that their mod­el describes a phase tran­si­tion from an ini­tial polarised state to a depo­larised state as a func­tion of increas­ing social influ­ence. The nature of this tran­si­tion depends on the dis­par­i­ty of ini­tial opin­ions: opin­ions that diverge strong­ly at the out­set trig­ger a so-called first-order (or explo­sive) depo­lar­i­sa­tion towards con­sen­sus, while opin­ions that are more cor­re­lat­ed to begin with lead to a sec­ond-order (or con­tin­u­ous) transition.

To test their mod­el, the researchers used data on cor­re­lat­ed top­ics – such as abor­tion and reli­gion – and uncor­re­lat­ed top­ics – for exam­ple, immi­gra­tion and mil­i­tary diplo­ma­cy in the Unit­ed States – from the Amer­i­can Nation­al Elec­tion Stud­ies. They found that com­mu­ni­ties asked to give their opin­ion on these sub­jects under­went a phase tran­si­tion from polar­i­sa­tion to depo­lar­i­sa­tion in numer­i­cal sim­u­la­tions of the mod­el, with indi­vid­u­als in the com­mu­ni­ty inter­act­ing and influ­enc­ing each other.

They stud­ied the mod­el under “mean field” con­di­tions, mean­ing that each indi­vid­ual can inter­act with all the oth­er indi­vid­u­als. “Since opin­ions are described by angles, it was nat­ur­al for us to mod­el con­sen­sus for­ma­tion as the align­ment of agents’ ori­en­ta­tions,” explains Michele Starni­ni. “This type of phase cou­pling is inspired by the Kuramo­to mod­el and is real­is­tic for small groups. In future work, we will test our mod­el on large inter­act­ing groups, such as social networks.”

“Anoth­er inter­est­ing appli­ca­tion that we are look­ing for­ward to imple­ment­ing involves simul­ta­ne­ous­ly mea­sur­ing the opin­ions of indi­vid­u­als with respect to mul­ti­ple top­ics and their social inter­ac­tions, to test the mod­el in this more real­is­tic setting.”

Isabelle Dumé