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