Scientific American - 09.2019

September 2019, ScientificAmerican.com 59

5Ÿ®y®y

Initial credences

Stable, opposing beliefs within a group Cliquish arrangement with stable opposing beliefs due to conformity

Experiment results (successes out of 10)

¹ùàÊblue)

ïyåïD` ́

5Ā¹Êorange)

ïyåïD` ́

5Ÿ®y®y

Updated credences

6 5

4

2

7

5

5Ÿ®yåïyÈÀ 5 Ÿ®yåïyÈ÷ 5 Ÿ®yåïyÈñ 5 Ÿ®yåïyÈŽ 5 Ÿ®yåïyÈ‹ 5Ÿ®yåïyÈê

"¹Āyà ̈yÿy ̈

¹†ïàùåï

(faint lines)

y ̈ŸyÿyåD` ́

ÊorangeËŸåUyïïyàUùï`¹ ́†¹à®åï¹D` ́幆U ̈ùy` ̈ŸÕùy

'Uåyàÿyàå ÿŸmy ́`yåyy§yàå 0à¹ÈD‘D ́mŸåï 'Uåyàÿyàå ÿŸmy ́`yåyy§yàå

Initial credences Experiment results (successes out of 10) Updated credences

0à¹ÈD‘D ́mŸåï¹ ́ ̈Ăå›Dàyåàyåù ̈ïå

ï›DïåùÈȹàïD` ́Êblue)

5 3

7

8

3

6

4

9

5

3

5

ING AND EXPERIMENTING:
Individuals in these models start with some
random level of certainty, or credence, about
whether ac tion A or B is better. They then
take the action they prefer—that is, “experi-
ment.” Their outcomes provide evidence
about the success of these actions, which they
share with neighbors. All individuals update
their credences based on what they observe.

CONVERGENCE ON TRUE BELIEFS: Over time, the social connec tions in these models mean that groups of people come to a consensus about whether A or B is better.
As they gather and share evidence, they usually learn that the better action is, indeed, better. Someone trying the worse action, for instance, will see how much better their
neighbor is doing and switch. Sometimes, though, strings of misleading evidence will convince the entire group that the worse action is better.

POLARIZATION:If we add social trust or
conformity to these models, they may no
longer reach consensus. If each individual
trusts the evidence that comes from those
who share their beliefs, polarized camps
that only listen to those in their group form.
Ifeach individual seeks to conform their
actions with group members, good ideas
fail to spread between cliques.

EVIDENCE SEEKERS, OBSERVERS AND
PROPAGANDISTS: In some cases, propa-
gandists try to mislead a group of people about
åŸy ́àyåù ̈ïåÎ=yD ́ùåyï›yåy®¹my ̈åï¹ àyÈàyåy ́ïDåy﹆yÿŸmy ́yåyy§yàåĀ›¹‘Dï›yà
yÿŸmy ́yjD‘à¹ùȹ†¹UåyàÿyàåĀ›¹ùÈmDïy Uy ̈Ÿy†åUDåym¹ ́åyÿŸmy ́yjD ́mDÈà¹ÈDž
‘D ́mŸåïĀ›¹®Ÿå ̈yDmåï›y¹UåyàÿyàåÎ

BELIEF UPDATING WHEN SELECTIVE RESULTS ARE IN PLAY: Industrial propagandists shape public belief by selectively sharing only those results that happen
to spuriously support the worse action. This can mislead the public, even in cases when groups of evidence seekers converge to a consensus about the true belief.
5›ŸååïàDïy‘ƹàÈùU ̈ŸmŸåŸ ́†¹à®D ́ïD§yåDmÿD ́ïD‘y¹†ï›yŸ ́›yày ́ïàD ́m¹® ́yåå ¹†åŸy ́`àyåù ̈ïåï¹®Ÿå ̈yDmÎ