← all systems · see it in The System
Social contagion
'Social contagion involves behaviour, emotions, or conditions spreading spontaneously through a group or network' (encyclopedia); Levy and Nail (1993): the spread of affect, attitude or behaviour 'where the recipient does not perceive an intentional influence attempt on the part of the initiator'. Network science: 'a vast array of processes that spread across networks, notably social phenomena such as fads, political opinions, the adoption of new technologies, and financial decisions' (Ugander et al. 2012).
Properties published scienceour proposal
Sourced fields, and our own readings.
| born at | at least as old as schooling vertebrates ('a ubiquitous behavior among taxa'); as a studied idea, 1895 (Le Bon) and the term in 1939 (Blumer) |
| born precision | rung-level for the process; year for the study |
| boundary | Inside: a state (feeling, belief, behaviour) passing from member to member of a group or network, the channels that carry it, the ties it runs on and the thresholds at which members take it up. Outside: deliberate influence that the receiver perceives as such ('separate from the collective behaviour which results from a direct attempt to exert social influence'), command and compliance, and clusters made only by like choosing like (homophily). |
| nested level | A process carried by groups of animals and people: built from members, their ties and their channels; embedded in groups, crowds and social networks; studied by science as social contagion, network contagion and crowd psychology. |
| scale | social |
Measures (20) published science
| happiness influence window (characteristic clock) | strongest when a friend became happy within the past half year; falls as the time between exams grows |
| happy friend within a mile | +25 (95% CI 1 to 57) % probability of being happy |
| rhythmic applause window (one recording) | 12-25 s |
| clapping frequency in the synchronized phase | approximately half the starting frequency ratio |
| happy next-door neighbour | +34 (7 to 70) % |
| happy sibling within a mile | +14 (1 to 28) % |
| happy coresident spouse | +8 (0.2 to 16) % |
| Framingham network studied | 4739 people followed 1983-2003 |
| reach of happiness and cooperation | 3 degrees of separation |
| cooperation multiplier | each extra contribution tripled over the experiment |
| voting trial size | 61 million Facebook users |
| moral-emotional words and diffusion | +20 per added word (n = 563,312 messages) % |
| village trial redemption by targeting (multivitamins) | nominated friends 74.3 (951/1280); random 66.2 (940/1420); most ties 61.0 (744/1220) % of tickets |
| nomination targeting gain over random | 12.2 (95% CI 6.9 to 17.9) |
| Honduran village networks mapped | 24,702 people in 176 villages; 117 outcomes; 22-month package |
| village spillover reach | 2 degrees of separation |
| fish school size tracked | about 150 fish |
| diffusion chain length | 10 people per chain |
| implausible network effects before adjustment (acne odds ratio; height share) | acne OR 1.62 (0.91 to 2.89); headaches 1.47 (0.93 to 2.33); height 0.18 (0.15 to 0.26); all small and insignificant after adjusting for shared environment odds ratio / coefficient |
| rumour reach without threshold (Daley-Kendall, homogeneous mixing) | a finite fraction r_inf = 1 - exp(-(1 + lambda/alpha) r_inf) for any spreading rate lambda fraction informed |
Open contentions (4) open contention
Where sources or readings disagree. Both stay on the map.
- C-394 Do behaviours spread through social networks, or do similar people just cluster together?
booked: The map keeps both: evidence of real spread from trials, and the point that ordinary observations confuse spread with similarity.
the other reading: observational clusters are generically confounded with homophily and shared environment (Shalizi and Thomas; Cohen-Cole and Fletcher's 'contagious' acne and height) - C-395 Do ideas and behaviours spread mainly through weak or strong social ties?
booked: The map follows the view that it depends on what spreads: weak ties for simple things, strong ties for complex behaviours such as voting.
the other reading: one rule for all spread: weak ties (Granovetter 1973) or strong ties (Bond 2012) - C-396 Are the most connected people the best starting points for spreading something?
booked: The map keeps the mixed findings side by side: sometimes yes, sometimes no better than random.
the other reading: hubs are (or are not) the best seeds in general - C-397 Does crowd behaviour spread like a contagion, or come from shared norms and identity?
booked: The map treats contagion as one mechanism, and keeps emergent norms, like-minded gathering and shared identity alongside.
the other reading: crowd behaviour better explained by emergent norms, convergence of like-minded people, or shared social identity (Turner, McPhail, Reicher)
Cross-scale patterns seen here (10) our proposal
Shapes that recur when the same kind of problem is solved at different scales, and how each shows in Social contagion. Each links to every other system that shows it.
- coordination-before-mapping (structural)
Applause synchronises, stadium waves travel and schools evade with no map of the whole: each member uses 'simple, robust measures' of its neighbours or the sound of the hall. - runaway from an unstable balance (structural)
A network sits stable until a small shock crosses enough thresholds; then a global cascade runs ('large but rare cascades... triggered by small initial shocks'). Its form varies: power-law sizes when connectivity limits it, bimodal 'more extreme' instability when node stability does (Watts); a bista - same role, different builds (structural)
One job, passing a state from neighbour to neighbour, is done by light in fish schools and stadiums, by sound in concert halls, by text in feeds and by face-to-face ties in villages. - trade one property for another (structural)
Clustering buys reinforcement and pays in reach: 'For simple contagions, too much clustering means too few long ties... For complex contagions, too little clustering means too few wide bridges'. Contagions sort along the dial: idioms and news (simple) vs political hashtags and costly behaviours (com - who opens the door (structural)
The tie is the port, made by the pair before anything spreads: nearly all voting transmission ran between close friends with face-to-face ties; coworkers showed no happiness effect; village spillovers ran along mapped friendships. - holding / entrainment (partial)
The audience's rhythm sets each clapper's tick: clapping frequency drops to about half in the synchronised phase. - lightning (partial)
A panic or evasion wave: a held state released all at once, after which the group is not the same ('the suddenness of panics'; 'rapid waves, or cascades'). Not gated or capped in the sources read. - same shape, opposite sign (partial)
Exposure moves an attitude in both cases; across the party line it moves it away: 'Republicans who followed a liberal Twitter bot became substantially more conservative'. Moral contagion runs within groups and less between them. - templating (partial)
The epidemic model's classes reused for rumours ('Their role is exactly the same of the susceptible, infected, recovered agents'), with one term changed (stopping when contacts already know), and 'similar models' used for pathogens, knowledge and innovation. - the whole sets the local reference (partial)
Each member reads its own move against the group: a threshold on the fraction of contacts who have adopted; 'the person's impression of the universality of behavior'; social comparison.
Valence: what it trades, and with whom (14) published science
Each exchange with a neighbour, read from both sides: who gains, who pays, and what carries it.
- Vertebrate (Vertebrata) via sys:electromagnetic-field · Social contagion gain
- Emotional contagion via sys:mental-representation · Social contagion gain
- Online social network (Facebook News Feed, Twitter) via sys:written-record · Social contagion gain
- Human (Homo sapiens) via sys:electromagnetic-field · Social contagion gain
- Cumulative culture (cumulative cultural evolution) via sys:mental-representation · Social contagion gain
- Science (as a human social and cultural system) via sys:mental-representation · Social contagion gain
- Complex network (scale-free growth, preferential attachment, small worlds; walked as an idea) via sys:complex-network · Social contagion gain
- Mass media via sys:written-record · Social contagion gain
- Human attention (the pool memes compete for) via sys:mental-representation · Social contagion mixed
- Human society (from local community up to societies) via sys:complex-network · Social contagion gain
- Social network via sys:complex-network · Social contagion gain
- Information warfare via sys:social-network · Social contagion even
- Public speaking (oratory) via sys:acoustic-wave · Social contagion even
- Stress (biology) via sys:sensory-environment · Social contagion mixed
Partners and parts published science
Built from / builds (3)
Sits inside / holds (4)
expressed by (6)
- Interoperability/Transduction (Bridging)
- Amplification/Scale-Up (Manufacture & Distribution)
- Sensing/Scanning
- Multi-Scale Data Fusion / Triangulation
- Social Learning / Public Information Use
- Quick Harm Avoidance
floor filled by (18)
- Vertebrate (Vertebrata)
- Emotional contagion
- Online social network (Facebook News Feed, Twitter)
- Acoustic wave (sound: pressure waves in matter)
- Human (Homo sapiens)
- Science (as a human social and cultural system)
- Social network
- Social ties
- Fish school (shoal)
- Complex contagion
- Weak tie (casual acquaintance)
- Crowd
- Human attention (the pool memes compete for)
- Echo chamber
- Human society (from local community up to societies)
- Information warfare
- Public speaking (oratory)
- Stress (biology)
represents (1)
Kinds (14)
- Emotional contagion
- Memetics (cultural selection)
- Behavioural contagion
- Simple contagion
- Complex contagion
- Disinhibitory contagion
- Echo contagion
- Hysterical contagion
- Financial contagion
- Moral contagion
- Rumour spreading
- Emergent norm theory
- Convergence theory
- Social identity approach to crowds
Notes published science
- Along ten-person retelling chains the risk message gets shorter and less accurate, but the feeling of risk travels with higher fidelity: the valence outlives the message (Moussaid et al. 2015).
- On Facebook, what predicts joining is how many separate groups of your friends have joined, not how many friends: once that is fixed, more friends predicts less joining (Ugander et al. 2012).
- Standard network methods found acne, headaches and even height 'contagious' between friends until shared surroundings were controlled for: a false alarm that keeps the field honest (Cohen-Cole and Fletcher 2008).
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
light: neighbours' movements, faces and crowds seen
- L1: Light from neighbours' bodies and movements: positions and body postures of fish in schools of about 150; a neighbouring section of stadium spectators leaping up with arms raised; a crowd seen by its members
- L2: Each receiver's visual field: 'individuals use simple, robust measures to assess behavioral changes in neighbors' (computed from tracked visual fields in golden shiners)
- L3: A threshold on what is seen: contagion in the school is 'complex, fractional'; in models of the stadium wave each spectator's summed neighbour influence, stronger from the side the wave comes from, is compared with an activation threshold spread over a range of values
- L4: The receiver's own move, seen in turn by its neighbours: a startle, or a leap up and sit down
- L5: The group wave: 'rapid waves, or cascades, of behavioral change' through the school; La Ola, which a critical mass must start; in Le Bon's crowd 'all emotions are very rapidly contagious, which explains the suddenness of panics'
mental representation: feelings, beliefs and norms carried from person to person
- L2: Affect carried: clusters of happy and unhappy people; each moral-emotional word raises diffusion by about 20%; emotions read in feeds shift the reader's own posts
- L4: Beliefs and judgements carried: perceived risk travels with higher fidelity than the message that carries it; knowledge spreads more readily than behaviour
- L5: Restraint and threshold in each mind: disinhibitory contagion (a wanted act held back by norms, released when others do it), echo contagion, hysterical contagion; reduction of restraints (Redl 1949, Wheeler 1966)
- L6: Emergent norms and shared identity: during milling key members suggest actions and followers fall in line; in mass emergencies a new shared social identity among strangers
- L7: The crowd's one direction: 'the turning by means of suggestion and contagion of feelings and ideas in an identical direction'; the crowd 'forms a single being'; echo chambers each around its own narrative
social network: the ties the spread runs on
- L1: Ties: friends, spouses, siblings, next-door neighbours and coworkers (4739 people in the Framingham network, 1983-2003); face-to-face networks of 24,702 people in 176 Honduran villages; 'close friends' and casual acquaintances
- L2: Neighbourhood shape: clustered-lattice versus random networks with the same people; the size of a contact neighbourhood versus its number of separate components
- L3: Bridges: long ties (weak ties, small-world shortcuts) carry simple contagions far; complex contagions need 'wide bridges' of several ties
- L4: Reach: up to three degrees of separation (happiness; cooperation); friends and friends of friends (voting); spillovers to two degrees (village health package)
- L5: States of the whole: large but rare global cascades in two regimes (connectivity-limited, power-law sizes; stability-limited, bimodal sizes); no epidemic threshold for simple spreading on scale-free networks; with group interactions a discontinuous transition and a bistable region
- L6: Where a spread starts: nominated friends of random villagers beat random and most-connected seeds; the most connected nodes trigger cascades only in the connectivity-limited regime
sound: voices, cheers and applause heard
- L1: Clapping and voices: applause after theatre and opera performances, recorded by microphones at a distance from the audience and close to single spectators
- L2: Each spectator hears the whole audience and near neighbours (the noise intensity 'heard at some distance from the audience or by one of the spectators')
- L3: Each clapper's period adjusts: clapping starts fast and 'reduces to approximately a half in the synchronized phase'
- L4: Rhythmic applause: the signal becomes periodic, the average intensity falls, and an order parameter shows 'the emergence of synchronization'
written record: posts, feeds and shared messages
- L1: Posts, messages and shares: status updates in the Facebook News Feed; political mobilization messages to 61 million Facebook users; 563,312 social media messages on three polarizing issues; photo reshares
- L2: Exposure: agents 'can only pay attention to a portion of the information they receive'; readymade share and retweet buttons 'automatically enable the spread of information as a simple contagion'
- L3: Counting the sources: political hashtags behave like complex contagions, idioms and memes like simple ones; on Facebook joining is 'tightly controlled by the number of connected components' in the contact neighbourhood
- L4: Reshare cascades: breadth early is the better sign of a large cascade; each echo chamber 'has its own cascade dynamics'; 'a few memes go viral while most do not'
- L5: Filters and bounds: withholding emotional posts from feeds shifted what readers posted; moral contagion 'bounded by group membership'; following a bot of the other side pushed Republicans further to their own side