We predict a riot: inequity, relative

royalsocietypublishing.org/journal/rspb
deprivation and collective destruction in
the laboratory
Guillaume Dezecache1,2,†, James M. Allen1,†, Jorina von Zimmermann1 and
Research Daniel C. Richardson1
Cite this article: Dezecache G, Allen JM, von 1
Department of Experimental Psychology, University College London, London, UK
2
Zimmermann J, Richardson DC. 2021 We Université Clermont Auvergne, CNRS, LAPSCO, Clermont-Ferrand, France
predict a riot: inequity, relative deprivation and GD, 0000-0002-9366-6287
collective destruction in the laboratory.
Proc. R. Soc. B 288: 20203091. Riots are unpredictable and dangerous. Our understanding of the factors
that cause riots is based on correlational observations of population data,
Downloaded from https://royalsocietypublishing.org/ on 28 January 2025

https://doi.org/10.1098/rspb.2020.3091
or post hoc introspection of individuals. To complement these accounts, we
developed innovative experimental techniques, investigated the psychologi-
cal factors of rioting and explored their consequences with agent-based
Received: 11 December 2020 simulations. We created a game, ‘Parklife’, that physically co-present partici-
pants played using smartphones. In two teams, participants tapped on their
Accepted: 31 August 2021
screen to grow trees and flowerbeds on separate but adjacent virtual parks.
Participants could also tap to vandalize the other team’s park. In some
conditions, we surreptitiously introduced inequity between the teams so
that one (the disadvantaged team) had to tap more for each reward. The
Subject Category: experience of inequity caused the disadvantaged team to engage in more
Neuroscience and cognition destruction, and to report higher relative deprivation and frustration.
Agent-based models suggested that acts of destruction were driven by the
interaction between individual level of frustration and the team’s behaviour.
Subject Areas:
Our results provide insights into the psychological mechanisms underlying
cognition collective action.

Keywords:
riots, relative deprivation, social identification,
collective action, Parklife
1. Introduction
Riots—defined as a ‘violent demonstration or clash of more than 100 citizens
involving the use of physical force’ [1, p. 4]—have long been a central topic
Author for correspondence:
in the social sciences and public debates owing to their societal consequences.
Daniel C. Richardson In the London riots of 2011, five people died, many more were injured and
e-mail: [email protected] property damage worth more than £200 million [2].
Why do people engage in riots? Politicians may caricature rioters as criminal-
minded individuals [3], but scientific investigation has shown the reasons to be
varied. Perhaps the leading psychological explanation for riots has been relative
deprivation theory (RDT) [4]. RDT stated that when people perceive a difference
between what they have and what they believe they deserve, they feel relatively
deprived. As the perceived disparity grows, so do frustration and resentment,
increasing the likelihood for engagement in collective violence. This theory has
contributed to explaining puzzling cases of collective violence, such as the
bread riots of the eighteenth century, in which people may not have suffered
from starvation, but engaged in collective violence, possibly because what
drove them to act was the relative (i.e. when compared with other groups in
† the society) and not absolute level of deprivation [4].
These authors contributed equally to this
Though popular, RDT has been subjected to much criticism [5,6]. Sociologi-
study.
cal work has failed to reveal a clear association between deprivation indices
(such as economic deprivation) and riots [6,7]. For example, economically
Electronic supplementary material is available deprived neighbourhoods were not more riotous than others during the
online at https://doi.org/10.6084/m9.figshare. 1960s racial disorders in the USA [8]. Economic deprivation is not the same
c.5611050. as relative deprivation (the latter is subjective and not necessarily linked to

© 2021 The Author(s) Published by the Royal Society. All rights reserved.
 Parklife game 2
blue team park red team park
time

royalsocietypublishing.org/journal/rspb
vandalized work
familiarization towards survey questions
feature
vandalizing

park
features work
towards
building

team assignment

tap to switch to
work building
switch to
vandalizing

Proc. R. Soc. B 288: 20203091

participants
participants vandalizing
building
Downloaded from https://royalsocietypublishing.org/ on 28 January 2025

Figure 1. Schematic of a Parklife experiment. After being familiarized with the Hive interface (familiarization), participants were placed into teams at random (team
assignment). They played a Parklife game for 3 min (central screen), answered a series of questions (survey questions) and then played a second game with a new
series of questions. (Online version in colour.)

economic hardship), but the idea that relative deprivation are part of the same group as others) plays a major role
cannot explain the emergence of riots has since been a in the emergence of riots. Riots emerge through changes in
major assumption in the field. social identification, beliefs that there is a shared problem
These criticisms, however, are themselves limited by a and that collective action can prove efficient to change
conceptual flaw and the practical problem of directly measur- things [3,18,19].
ing relative deprivation and riot participation. Past tests used In this work, we sought to experimentally examine the
imperfect proxies for relative deprivation, such as aggregate causal role of relative deprivation in the emergence of riot-like
(rather than individual) and objective (rather than subjective) phenomena, as well as to measure the importance of social
measures of deprivation [9]. Imperfect proxies for rioting identification in the making of collective action. We developed
were also used, such as occurrences of disorders in a given a laboratory-based experiment to identify causal factors, and
neighbourhood, a measure which cannot be linked to specific modelled their interactions in agent-based simulations.
individuals whose level of relative deprivation is known. The experiment used a group interactive computer game,
People’s reported willingness to participate in a riot was Parklife, that large numbers of physically co-present partici-
measured, rather than actual participation (e.g. [10,11]). pants can play simultaneously using smartphones or tablets.
Even though there is evidence that relative deprivation (and In contrast with many games used in psychology, Parklife can
particularly, group or fraternal relative deprivation or the be played in the physical presence of other participants. Parklife
extent to which one feels she belongs to a deprived group resembles popular ‘time management’ apps in which players
[12]) can be associated with willingness to join protests [13] expend effort to develop resources, and are rewarded by a
or endorse violent actions [14], it has not been shown that growing world. Participants are randomly placed into two
it causes people to engage in collective destruction. teams. By varying the comparative effort to create ‘park
In recent literature, RDT and its relationship with hostile features’, we induce feelings of relative deprivation in the dis-
aggression has sparked new empirical study [12]. For advantaged team. We, therefore, have two conditions: the
example, Greitemeyer & Sagioglou [15,16] found that partici- equal game with no difference between the teams, and
pants told that they were of lower socio-economic status the unequal game where the disadvantaged team must work
(SES) behaved more aggressively than participants told they harder for equal rewards. By allowing (and measuring) vandal-
were of higher SES. At the group level, only one study ism of the other park across teams (advantaged and
has—to our knowledge—directly investigated the relation- disadvantaged) and conditions (equal and unequal), we
ship between the experience of inequity and collective tested the effect of relative deprivation on vandalism.
hostile aggression. Abbink et al. [17] showed that small More specifically, in the experiment, participants are
groups of participants treated with inequity coordinated to invited to the laboratory and to sit around a table. They are
deprive the favoured groups of their earnings. In this case, randomly assigned to two teams, and two empty parks are
however, the outcome of the violent behaviour is instrumen- displayed on a large shared screen, visible to all (figure 1).
tal (i.e. to reduce others’ earnings) and may not correspond to By tapping on a circle on their device, participants contribute
core motivations for hostile aggression. Additionally, feelings towards their team’s total work, which is displayed by a rising
of relative deprivation were not measured. This is needed to bar at the side of the park. Once a second, the average number
draw conclusions about any relationship between relative of team members working is calculated (thereby reducing the
deprivation and destructive collective action. As such, a impact of different team sizes), and added to the team’s bar
direct test of the association between the subjective feeling (visible to the side of each team’s park). It takes 4 units of
of relative deprivation and actual participation in collective work per team for the bar to reach the top, at which point a
violence is, to our knowledge, missing. park feature (e.g. a bench or flower bed) is built. If all members
Besides relative deprivation and associated feelings of of a team are working, it thus takes 4 s to build a feature.
frustration, social identification (i.e. how much we feel we Participants also have the option of switching their efforts
 from ‘do’ to ‘undo’. Tapping ‘undo’ contributes towards a world. However, we hypothesize that the psychological mech- 3
second bar on the screen. When the undo bar has filled, a fea- anisms at work are the same. Unlike other studies on relative

royalsocietypublishing.org/journal/rspb
ture in the park of the other team is vandalized, appearing on deprivation and collective violence, our paradigm measures
screen to be broken, and as before it takes 4 units of vandalism actual, direct and face-to-face collective destructive efforts
to destroy the park of the other team. A third option is available (rather than intention to join a protest, or individual aggressive
to players, that is, ‘to do nothing’. As the aim of the game is to behaviour), and unlike other studies of inter-group compe-
build as many park features as possible, regardless of the state tition (e.g. [21]), our study allows groups to act in real time
of the other park (as announced by the instructions), switching in a more realistic scenario. Such collective destruction
to ‘undo’ or to not tap at all seem irrational. As the state of the would be despite incentives to maintain collective constructive
other team’s park has no direct bearing on a team’s success, efforts, or the effortless option to just ‘do nothing’.
participants are taking away effort from improving their own
team’s park when tapping ‘undo’ or doing nothing. Aggressive
acts are operationalized as the individual and collective
decisions to tap ‘undo’ rather than ‘do’ or just staying idle. 2. Results

Proc. R. Soc. B 288: 20203091

In each session, participants played two games of Park- We ran 19 experimental sessions with a total of 203 partici-
life, each 3 min long. In one of the games, the two teams pants. Internet connectivity issues caused some data to be
had to do equal amounts of work to be rewarded with a excluded, if, for example, the server could not assign the par-
Downloaded from https://royalsocietypublishing.org/ on 28 January 2025

park feature (the equal game). In the other game, the unequal ticipant to a team, the participant was not active for the whole
game, one of the teams had to tap twice as much to be duration of the session, or if they had to reconnect to the
rewarded with each feature (i.e. the disadvantaged team server. This left us with 171 participants across 19 sessions
must now produce 8 units of work to build a feature), thereby with full data.
producing an inequity of reward between the teams. This In the figures throughout, the disadvantaged team’s
inequity was not announced to participants. We reasoned actions are shown in red and the advantaged team in blue
that this structural difference in the game (which participants (though in the experiment, team colours were counterba-
experienced as they play) would induce feelings of relative lanced). For the equal games (when the teams are equitably
deprivation in the disadvantaged team and cause its mem- rewarded for their work), the teams are shown in grey.
bers to engage in more acts of vandalism against the Figure 2 shows the actions that the teams took during the
opposing team when compared with the other (advantaged) games, and the consequences they had for the parks, in terms
team. Note also that the costs of vandalism remain equal in of park features that were built and destroyed. As we pre-
both teams, irrespective of the type of games (unequal or dicted, in the so-called unequal games (when one team has
equal). Although less costly than building, the vandalism to work more than the other to produce a park feature),
option remained more costly than staying idle, a low cost the disadvantaged teams built fewer park features and
and non-aggressive option, explicitly available to all players. vandalized more of the other team’s park features.
For players on the disadvantaged team, their efforts to To test our hypothesis on the link between inequity, rela-
build a park return fewer rewards. Switching to vandalism tive deprivation and collective destruction, we analysed
might be a way to affect a change in the game with lower individual’s actions during the course of the games. Our
effort. But crucially, there are different outcomes too. Players’ key dependent variable was the vandalism rate: the number
switching to vandalizing make the choice to abandon the of participants’ taps to undo (i.e. to vandalize the other
alternative (the more effortful action of building the park), team’s park) as a proportion of the total number of taps
and instead decide to engage in lower effort, anti-social that they made in the game. We analysed the vandalism
action against the other team (even though that indirectly rate as a function of game equality (equal versus unequal),
harms their own park too). The effort difference is a feature team membership (whether the individual was in the disad-
of our paradigm that may parallel real-world situations. The vantaged versus advantaged team for the unequal game),
conditions under which participants make that sort of choice and game order (whether they played the unequal game
(abandoning one type of effortful action to invest in anti- first or second). Figure 3 shows the observed vandalism
social actions ) are precisely what we want to investigate here. rates for the advantaged and disadvantaged participants, in
After each game, we privately polled participants about equal and unequal games.
their emotional state, their feelings towards each other, the We used Bayesian mixed models to quantify the evidence
opposing team and the game that they played. We counterba- that each of our experimental factors influenced participants’
lanced several game features between groups such as the vandalism rate. Mixed models are able to account for the
order of the equal and unequal games and whether the effect of individual participants being nested in a particular
disadvantaged team was red or blue. group, and the Bayesian approach avoids some of the pro-
Our experimental paradigm captures a key aspect of riots: blems associated with null hypothesis testing [22], or the
they are a complex emergent phenomenon. We then employed need for a formal ‘stopping rule’. Our models employed
agent-based models to provide insight into complex behav- weakly informative priors that were scaled following the
iour that unfolds over time [20]. These models tested a standard procedures (for full details and model specification,
number of hypotheses related to the central mechanisms see below).
that drive emergent phenomena in collective behaviour— In figure 3, to the right of the observed data, we show the
frustration and relative deprivation, social identification and distributions of the estimated differences between our exper-
social norms—and to see how they interact with inequity imental conditions. The percentage of this distribution that is
and relative deprivation produced by our game. greater than zero is known as the maximum probability of
Like many other psychology experiments, the experience effect (MPE), which directly quantifies the probability that
of our study is not exactly like experiences in the everyday the manipulation condition had an effect on behaviour. We
 advantaged equal disadvantaged 4

royalsocietypublishing.org/journal/rspb
(a) (b)
work actions park features built

10
0.6

0.4 0

Proc. R. Soc. B 288: 20203091

vandalize actions 1.2 park features vandalized
Downloaded from https://royalsocietypublishing.org/ on 28 January 2025

0.8
0.2

0.4

0 0

0 50 000 100 000 150 000 0 50 000 100 000 150 000

game time (ms) game time (ms)

Figure 2. Averaged time-course of the probability of actions taken by participants (a) and the number of park features affected as a consequence (b) for advantaged
and disadvantaged teams. At the top are work actions and features built, on the bottom are vandalize actions and features vandalized. (Online version in colour.)

(a)
0.2
advantaged
1.5
equal 0.1
estimated difference between equality conditions
1.0 0

–0.1
0.5
–0.2
density

0 0.25 0.50 0.75 1.00

(b) 0.2
disadvantaged
1.5 equal 0.1

1.0 0

0.5 –0.1

–0.2
0 0.25 0.50 0.75 1.00 density
vandalism rate
Figure 3. Distributions of the observed vandalism rates for the advantaged (a) and disadvantaged (b) participants, when they were playing in the equal (grey) and
unequal (coloured) games. Mean rates in dotted lines. On the right are distributions of Bayesian estimations of the differences between equal and unequal games,
with 95% of distribution shaded. The percentage of this distribution that is greater than zero is known as the MPE, which directly quantifies the probability that the
manipulation condition had an effect on behaviour. Here, it shows that there was only strong evidence of a difference between games for the disadvantaged teams.
(Online version in colour.)
 report below MPEs for each of our experimental factors and number of others across different teams in continuous time, 5
contrasts within levels. The Bayesian approach favours quan- we build an agent-based model (ABM) [24,25]. The model’s

royalsocietypublishing.org/journal/rspb
tifying the strength of evidence in this way, rather than goal was to evaluate two key hypotheses: individuals
simply reporting whether or not an (arbitrary) threshold of become frustrated through comparisons with better placed
significance has been passed. Having said that, researchers individuals or groups; also, and social identification and
generally suggest that an MPE of above 90% or 95% can be norms of behaviour play a role in escalating cycles of conflict.
thought of as ‘strong evidence’ [23]. In figure 3, we can see Our ABM was designed to quantify—in the language of Park-
a grey area that corresponds to 95% of the estimate distri- life—whether individuals vandalize at random, or whether it is
bution. When this interval does not cover zero, it can be owing to social comparison and frustration, the creation of
seen as strong evidence for a difference between conditions. norms in each team, or some combination of factors.
In addition to these Bayesian analyses, we ran frequentist To summarize the model’s conclusions, both relative depri-
analysis, which produced a corresponding pattern of results vation (in the form of park differences) and other participant’s
(see the electronic supplementary material). behaviour are key to the increased proportion of vandalism in
As predicted, there was an increase in vandalism rates for the disadvantaged team in the unequal condition (more details

Proc. R. Soc. B 288: 20203091

the disadvantaged teams when they were in the unequal may be found in the electronic supplementary material). We
game compared to the game where they were treated equita- find there is a bias towards information from a participant’s
bly (MPE = 99.7%). This was not the case for the advantaged own team, and that participants engage in coordinated behav-
Downloaded from https://royalsocietypublishing.org/ on 28 January 2025

teams, where there was no evidence of a difference between iour, distributing vandalism and work across the participants
game types (MPE = 62.1%). There was evidence of a main within their own team. Alongside this, we find positive evi-
effect of a higher amount of vandalism overall in the unequal dence that those in the disadvantaged condition are not
games compared to the equal ones (MPE = 96.3%), but no evi- behaving rationally, i.e. only vandalizing through boredom
dence of a higher amount of vandalism between the teams or to make something happen, and instead are responding
across all game types (MPE = 52.2%). There was no evidence to both park differences and the behaviour of the other
for more vandalism when unequal came first (MPE = 70.4%), participants within the game.
and no other factors had a significant impact on the Our model is designed to mimic Parklife as closely as poss-
proportion of vandalism across conditions. ible: each individual within the model is placed in a team in
As we were interested in how the experience of inequity either the equal or unequal condition, plays the game for 180 s
changed participants’ experience, they privately reported and can work or vandalize. The model agents have access to
how they felt about the game and each other after both the the same information as the participants in Parklife, i.e. the
equal and unequal game, on a number of custom-made items number of features in the parks, and the number in each team
(figure 4, left). To quantify their response to inequity, we sub- working or vandalizing at any one time. Agents may decide
tracted their answers following the equal game from the to work or vandalize based on park differences or the behaviour
answers following the unequal game, and compared these of others. This information is taken from both teams, and which
difference scores between advantaged and disadvantaged team agents focus on when deciding whether to work or vanda-
teams. We ran a Bayesian mixed model for each item, with lize is biased (the details of which are described in the ESM).
the factors of team and game order, nested in the experimental Finally, the output of the model is the proportion of agents on
group. From these runs, we generated distributions for the esti- each team vandalizing in each second.
mates of the difference scores (figure 4, centre) and calculated We conclude that participants spend the majority of their
MPEs for each. time focusing on the state and behaviour of their own team.
Inequity had a very different effect on participants’ explicit However, over time, park differences increase, and so on the
ratings depending on whether they were advantaged or disad- minority occasions that cross park comparisons are made,
vantaged. For all questions, MPEs were greater than 99.9%, those in the unequal, disadvantaged condition become fru-
except the two questions that asked if participants would act strated, and vandalize. In the full model, our findings
pro-socially towards the other team by sharing a reward suggest that players focus mainly on their own team, and coor-
(MPE = 80.8%) or build a bridge between the parks (MPE = dinate their behaviour by performing the opposite function of
90.3%). Critically, the disadvantaged teams felt that they those on their team (i.e. if many team mates are working, they
deserved more park features compared to what they received vandalize and vice versa). Participants balance team behaviour
and that their team was unfairly disadvantaged, suggesting between working and vandalizing, providing evidence of
they experienced a shared fate with others. They also reported coordinated behaviour across the teams. Examples of 10 runs
higher frustration, anger and resentment in unequal versus at our fitted parameters are shown in figure 5, and compared
equal games when compared with the advantaged teams. to our empirical data.
Interestingly, items linked to social identification (whether To further test the motivations and mechanisms for vand-
people felt their team demonstrated its strength and identifi- alism in Parklife, we ran two simpler versions of the model:
cation with the team) were higher in advantaged than (the frustration-only and asocial models) in which individuals
disadvantaged teams. do not pay attention to others’ behaviour. In the frustration-
only model, we remove the importance of social norms. The
asocial model is designed to test if participants were simply
tapping randomly or performing a cost–benefit analysis in
3. Computational model keeping park differences to a minimum: if the latter, those
Our experimental work identified in the unequal game a set of in the disadvantaged team in the unequal game would
causal factors that produced acts of collective destruction. To choose to vandalize rather than work with a higher prob-
understand how these factors may interact, we used compu- ability, as this decision reduces the effort to change the state
tational modelling. As individuals can interact with a of a park/reduce park differences.
 disadvantaged unequal game estimate of team difference 6
advantaged equal game between unequal–equal

royalsocietypublishing.org/journal/rspb
disagree agree

I feel happy

I feel frustrated

I feel relaxed

I enjoyed playing this game

I felt attached to our park

Proc. R. Soc. B 288: 20203091

I feel angry and resentful towards my team
I feel angry and resentful towards the other
team
Downloaded from https://royalsocietypublishing.org/ on 28 January 2025

I feel angry and resentful towards the

experimenter
my team deserved more park features
my team was responsible for the outcome
of the game
this game was fair

my team performed well in this game

I identified with my team

my team demonstrated its strength

I felt in control of my actions during this
game
my teammates acted exactly like I did

I want to play this game again

compared to the other team, my team was
unfairly disadvantaged
Would you like to build a bridge between
the two parks to connect them and share
the park features built?
If your team earned an extra monetary
reward would you be willing to share?
Figure 4. Responses to post-game survey items in the equal (circle) and unequal (triangle) games. For example, the first item shows that (compared to the equal
game) in the unequal game, the disadvantaged teams were less ‘happy’, whereas the advantaged teams were more ‘happy’. To the right are Bayesian estimates of
differences between advantaged and disadvantaged teams in the size and direction of these shifts in response inequity. In the first item, 95% of this distribution
(shown in shaded area) does not include zero, and so there is strong evidence that there is a reliable difference between teams. (Online version in colour.)

Comparing our different models using Bayes factors, are using the catch-all term ‘frustration’ for the emotion
we find positive evidence for the full model over the driving vandalism within Parklife.
asocial model (Bayes factor = 9.4) and we, therefore,
conclude that individuals are behaving by neither tapping
at a base rate, or tapping at an increased constant rate
in only the unequal, disadvantaged condition in order to 4. Discussion
keep park differences to a minimum. We found that What are the psychological mechanisms that make people
there is also strong evidence for the full model over one the engage in violent collective behaviour? One view has been
frustration-only model (Bayes factor = 12.8), therefore that when people feel they have less than deserved, they
showing the importance of social norms and team behaviour experience relative deprivation. This causes frustration and
in Parklife. hostile aggression [4]. This explanation was questioned
Our use of the term ‘frustration’ is not a commitment to based on a lack of substantial empirical evidence [9,26].
any specific model of anti-social behaviour. Emotional Here, we used a laboratory-based experimental approach to
states are hard to identify, and, therefore, in this work, we investigate behavioural and psychological responses to
 disadvantaged advantaged 7
0.5

royalsocietypublishing.org/journal/rspb
0.4
0.3

equal
mean proportion of vandalism
0.2
0.1
0
Parklife data
0.5
simulation
0.4

unequal
0.3

Proc. R. Soc. B 288: 20203091

0.2
0.1
Downloaded from https://royalsocietypublishing.org/ on 28 January 2025

0
0 50 100 150 0 50 100 150
time
Figure 5. Ten example runs of the model using Maximum A Posteriori point estimates of the posteriors for each team and game condition (blue faded lines are a
single simulation, red solid lines are the empirical data, both averaged over 19 runs). (Online version in colour.)

inequity, an approach which allowed us to directly measure be necessary for collective action to occur. In Parklife, partici-
actual collective destructive behaviour in response to per- pants coordinate (as our models show), but it remains open
ceived inequity. We found that the experience of being whether they also explicitly understand that they are taking
treated with inequity can lead to acts of collective aggression part in a collective action. The responses to the post-game
in a disadvantaged group, associated with reports of being questionnaire suggest they do as, for example members of
unfairly treated together with one’s own team. In our exper- the disadvantaged team felt their team was unfairly disad-
iment, hostile behaviour took the form of damaging another vantaged in unequal games, with a reduction of this belief
team’s park. This behaviour was also detrimental to the indi- in the advantaged team. This should be more directly
viduals themselves, as they were spending time vandalizing addressed using additional post-game self-report items.
the opposition rather than improving their own park, or Of course, tapping a given button in a virtual game is not
simply doing nothing. This suggests that these acts of collec- equivalent to deciding to riot in the real world. It was not
tive destruction were not a cold, purely rational strategy the goal of Parklife or our models to accurately simulate a
to succeed at the task. Indeed, violent responses were associ- real riot. Although the stakes and context are different, we
ated with feelings of frustration, deprivation and of being contend that the same psychological mechanisms that
treated unfairly. turn frustration into violence are at work. Experimental
Agent-based modelling confirmed this, and added approaches such as these can help investigate social beha-
further insight into how these psychological factors interact viours that are difficult to study in the real world with
with additional factors associated with the emergence of precision or control. Our approach offers a number of advan-
riots in the literature, such as social identification and tages over other non-social and non-realistic approaches.
the importance of group norms [18]. Our best fitting model Participants in Parklife meet with real participants and inter-
shows that, while frustration is a key explanatory variable act with them during the game; what is more, they are meant
for the increase in vandalism in disadvantaged groups, to produce a virtual but perceivable object (a park), rather
social identification and other’s behaviour play an important than allocating virtual resources they may have great difficul-
role. In this model, participants begin focused on their own ties representing. Our approach thus combines the benefits
parks, with low initial frustration. However, on the relatively of tightly controlled experimental methods, with the
rare occasions when the disadvantaged group compares advantages of realism, physical co-presence and interaction.
themselves to the advantaged group, their frustration and Psychologists have learnt much about group processes, for
aggression levels spike; but individuals continue to act in example, using methods such as the minimal group paradigm.
relation to their team mates, ensuring that labour is distribu- In typical experiments, participants are assigned to a group by
ted among vandalizing and working. This interaction an arbitrary or random criteria, and then asked to allocate
between individual frustration and team behaviour captures abstract resources to in- and out-group members. Real-life
our experimental data better than frustration alone. The social identities are not formed like this, of course, and in
mechanisms demonstrated here may have wider ramifica- real life, we do not allocate resources to each other like that.
tions for the study of collective behaviour, and in particular, Yet, the minimal group paradigm turns on the same psycho-
the effect that through coordinated action in a group setting, logical mechanisms that operate in the real world, and so
division of labour emerges naturally across each team. provides insight into real-world behaviour. Similarly, the
One important distinction in the literature has been goal of Parklife is not to recreate the circumstances of a riot,
between individual (or personal) and collective (or fraternal) but to create a game that turns on the same mechanisms of
relative deprivation [12], with the finding that the latter may social identification and relative deprivation.
 Our results add to the evidence that essentialism—the observed historically at the population level, and we under- 8
notion that riotous crowds are simply made up of violent stand little of the psychological responses to inequity at the

royalsocietypublishing.org/journal/rspb
people—is an inadequate explanation. Because people were behavioural level. Our experiment and simulations provide
randomly assigned to experimental conditions, collective evidence for one piece of this puzzle, showing that the experi-
violence can be produced by the situation alone. ence of social inequity and the behaviour of others make
The game mimicked situations in which contributing to people more likely to engage in acts of collective aggression.
building an item is more costly than destroying it (the disad-
vantaged team had to work twice more to build, but
vandalizing was equally costly between the teams). This
asymmetry could contribute to the finding that members of
5. Methods
the disadvantaged team prefer to vandalize in unequal (a) Code
games. Evidence from both the fact that participants choose The code to reproduce the analyses can be found at: https://osf.
to vandalize rather than do nothing, and our modelling led io/agbc3/.
us to argue that the asymmetry between the two teams, intro-

Proc. R. Soc. B 288: 20203091

duced by contributing to building an item being more costly
than destroying it (the disadvantaged team had to work
(b) Subjects
Because we were to employ Bayesian analysis (that does not
twice more to build, but vandalizing was equally costly require a pre-determined sample size or articulated stopping
Downloaded from https://royalsocietypublishing.org/ on 28 January 2025

between the teams), results in relative deprivation, and it is rule), we collected as much data as we could within a specific
this process which results in the increased vandalism. Of time window. Our sample size has, therefore, not been pre-regis-
course, ruling out all processes through which this may tered. We tested 203 participants (129 females) in 19 groups.
occur is difficult, and any change in reward structure Participants were recruited from the SONA system of the Univer-
would result in an asymmetry of costs, introducing its own sity College London. They were between 18 and 55 years old
possibilities. This said, it is a necessary step to experimentally (M = 20.97, s.d. = 4.57). They were compensated 5 GBP for their
test this in the future, systematically and independently vary- participation, or given course credit. Our goal was to run partici-
ing the relative costs and benefits of building and destroying pants in groups of 10. However, because it was challenging to
recruit and ensure the attendance of exactly the same number
in a fully factorial design, to test its impact on the observed
of participants each session, we ran opportunistically with who-
individual and collective behaviour.
ever came to each session, resulting in a range of group sizes
Social identification, social membership and histories are from 4 to 23 (M = 9.63, s.d. = 6.28). Pilot work suggested that
indisputable in the making of real-world riots, and riots within this range, group size did not have a systematic effect
involve groups that are already socially structured, or at on vandalism rates. Internet connectivity issues caused some
least, circumstances that favour shared identities and social data to be excluded. This left us with 171 participants across 19
norms [27,28]. Yet, minimal groups of randomly assigned sessions with full data for analysis.
teams have enough shared sense of identity to perceive
their group as being treated unfairly and respond with
aggression. In future work, we will explore how teams with
(c) Procedure
Before taking part in the experiment, subjects were asked to fill in
different social membership and histories will respond
an online questionnaire. This questionnaire measured a number
differently to the experience of inequity. of psychological traits. Upon arrival, subjects were seated in a
Riots are a paradigmatic example of emerging collective room, around a table and were asked to fill in a participation con-
behaviour, and as such, may be seen as a form of collective sent form. On a voluntary basis, participants could be equipped
action. Although we have not explicitly differentiated with a wristband to measure physiological indices (data to be
between individual- or group-level relative deprivation [12] reported elsewhere). Participants were instructed to join the web-
within Parklife, we observed the emergence of group-level site ‘thehive.sc’ with their smartphone or a tablet we provided.
behaviour through the interactions of individuals. Although On this website, they provided a subject number we assigned
decisions are made on the individual level, information is to them, as well as basic demographic information (gender and
received on the group level, as only information on the geographical area). Participants were then presented with a dot
on their screen device that they could drag around with a
parks and the total working or vandalizing on each team is
finger. We directed their attention to a central display which
available. Within a dynamic group game such as Parklife, it
showed the dots of all participants moving in real time.
is, therefore, important to consider behaviour across all Participants were then randomly assigned to two teams,
levels. Through this, our results may have broader impor- which was indicated by the dots on their device and onscreen
tance for group behaviour in other circumstances, both in changing colour. When the subjects were in odd numbers, one
humans and in other species. We find that coordinated be- team received one more player but participants were all told
haviour emerges from an interaction of competition and this would not affect the game outcome as work effort is
cooperation, and that this is true even with relatively weak scaled by the system to accommodate different team numbers.
social identification (teams are only allocated at random, They were then instructed that they would play two games of
and no existing social identities are considered). When indi- Parklife, each lasting about 3 min, and would answer some ques-
viduals are placed in a group environment with few tions afterwards. They were asked not to speak during the game.
The interface (figure 1) was explained to them, and they tried it
restrictions, we still observe collective group behaviour.
out in a 1 min practice game while we pointed out the infor-
Finally, this work demonstrates the importance of inequity
mation onscreen, and how the bars indicated their team’s
in the emergence of coordinated behaviour. current actions (figure 1). By the end of the practice game, we
Unequal allocations have societal consequences [29]. In ensured that participants understood how to play.
addition to the economic and practical difficulties caused Participants were asked not to speak during the game. They
by poverty, there might be a pervasive psychological response could see each other and potentially communicate non-verbally.
to real and perceived inequity. These correlations have been Because teams were assigned randomly, they did not know
 which other participants were on their team. Explicit communi- specified as: 9
cation or planning within a team was difficult, and we did not
observe any attempts. vandalism rate
team
game equality
game order

royalsocietypublishing.org/journal/rspb
Participants then played two Parklife games—one equal and þ (1 jgroup) þ (1jparticipant):
one unequal (order counterbalanced between experimental ses-
sions), together with whether the red or the blue team were The model had an explanatory power of around 53.29%
advantaged or disadvantaged. In the unequal game, the (median absolute deviance (MAD) = 0.043, 95% confidence
amount of presses on the ‘do’ button necessary to generate a interval (CI) = (0.44, 0.61), adjusted R 2 = 0.29). The electronic sup-
park feature was two times higher in one group than the other. plementary material gives the full parameter estimates of the
This was not announced to the participants. The amount of model with median, MAD, 95% CI (CI− CI+), MPE and overlap
presses on the ‘undo’ button to dismantle a feature in the park for each term.
of the other team remained equal between groups. Between In addition to these Bayesian analyses, we ran frequentist
games, participants were randomly reassigned to red and blue analysis using more conventional mixed models. These produced
teams, in an attempt to reduce carry over effects from one a corresponding pattern of results and can be seen in the
game to the next. Following each game was a set of custom- electronic supplementary material.

Proc. R. Soc. B 288: 20203091

made survey questions (figure 4). Participants indicated their
Ethics. We obtained ethical approval from the UCL Research Ethics
agreement with statements by moving a dot on their screen
Committee (approval ID number: 3828/003).
across a Likert scale (figure 1). Finally, participants were asked
Data accessibility. The code to reproduce the analyses can be found at:
debriefing questions, told the aims of our study and thanked
Downloaded from https://royalsocietypublishing.org/ on 28 January 2025

https://osf.io/agbc3/.
for their participation.
Authors’ contributions. G.D.: conceptualization, data curation, funding
acquisition, investigation, methodology, project administration, super-
vision, writing—original draft, writing—review and editing; J.M.A.:
(d) Statistical analysis conceptualization, data curation, formal analysis, investigation, meth-
Our mixed models used fixed effects for the participants team odology, project administration, resources, software, validation,
(advantaged/disadvantaged), the game equality (equal/ visualization, writing—original draft, writing—review and editing;
unequal) and game order (equal first/unequal first). There J.v.Z.: conceptualization, data curation, funding acquisition, investi-
were random effects for the experimental group and the partici- gation, methodology, project administration, supervision, writing—
pant, with random intercepts. We used R (v. 3.4.3 [30]) and the review and editing; D.C.R.: conceptualization, data curation, formal
package rstanarm (v. 2.18 [31]), employing weakly informative analysis, funding acquisition, investigation, methodology, project
priors that were scaled following the standard rstanarm pro- administration, resources, software, supervision, validation,
visualization, writing—original draft, writing—review and editing.
cedure (full priors are reported in the electronic supplementary
All authors gave final approval for publication and agreed to be held
material). From 4000 samples, we generated estimates of the pos- accountable for the work performed therein.
terior distributions of the model parameter coefficients, which
Competing interests. We declare we have no competing interests.
quantify the strength of the evidence that each experimental con-
Funding. This study has been funded by a grant from the Nuffield
dition influenced behaviour in a consistent way. Below we report
Foundation (The Psychological Roots of Societal Self Harm, 42868)
the estimates of the differences between experimental conditions, awarded to D.C.R., and a British Academy Newton International Fel-
using the package psycho (v. 0.3.7 [32]). lowship (NF 171514) and Alumni Follow-on Funding awarded to
We fitted a Markov chain Monte Carlo, details of which G.D. G.D. also acknowledges the support received from the Agence
can be found in the electronic supplementary material. Using Nationale de la Recherche of the French government through the
the formula notation in the R stats package, the full model was program ‘Investissements d’Avenir’ (16-IDEX-0001 CAP 20-25).

References
1. Braha D. 2012 Global civil unrest: contagion, self- 7. Miller AH, Bolce LH, Halligan M. 1977 The J-curve analytic review. Pers. Soc. Psychol. Rev. 16,
organization, and prediction. PLoS ONE 7, e48596. theory and the black urban riots: an empirical test 203–232. (doi:10.1177/1088868311430825)
(doi:10.1371/journal.pone.0048596) of progressive relative deprivation theory. Am. Polit. 13. Wright SC, Taylor DM, Moghaddam FM. 1990
2. Bencsik P. 2018 The non-financial costs of Sci. Rev. 71, 964–982. (doi:10.1017/S00030554 Responding to membership in a disadvantaged
violent public disturbances: emotional 00265180) group: from acceptance to collective protest.
responses to the 2011 riots in England. 8. Spilerman S. 1976 Structural characteristics of cities J. Pers. Soc. Psychol. 58, 994. (doi:10.1037/0022-
J. Hous. Econ. 40, 73–82. (doi:10.1016/j.jhe.2018. and the severity of racial disorders. Am. Sociol. Rev. 3514.58.6.994)
03.002) 41, 771–793. (doi:10.2307/2094726) 14. Obaidi M, Bergh R, Akrami N, Anjum G. 2019
3. Stott C, Drury J. 2017 Contemporary understanding 9. Brush SG. 1996 Dynamics of theory change in the Group-based relative deprivation explains
of riots: classical crowd psychology, ideology social sciences: relative deprivation and collective endorsement of extremism among western-born
and the social identity approach. Public violence. J. Confl. Resolut. 40, 523–545. (doi:10. Muslims. Psychol. Sci. 30, 596–605. (doi:10.1177/
Underst. Sci. 26, 2–14. (doi:10.1177/ 1177/0022002796040004001) 0956797619834879)
0963662516639872) 10. Muller EN. 1972 A test of a partial theory of 15. Greitemeyer T, Sagioglou C. 2016 Subjective
4. Gurr TR. 2015 Why men rebel. New York, NY: potential for political violence. Am. Polit. Sci. Rev. socioeconomic status causes aggression: a test of
Routledge. 66, 928–959. (doi:10.2307/1957487) the theory of social deprivation. J. Pers. Soc. Psychol.
5. Tilly C. 1971 Reviewed work: why men rebel by Ted 11. Walker L, Mann L. 1987 Unemployment, relative 111, 178. (doi:10.1037/pspi0000058)
Robert Gurr. J. Soc. Hist. 4, 416–420. (doi:10.1353/ deprivation, and social protest. Pers. Soc. Psychol. 16. Greitemeyer T, Sagioglou C. 2017 Increasing wealth
jsh/4.4.416) Bull. 13, 275–283. (doi:10.1177/014616728 inequality may increase interpersonal hostility: the
6. McPhail C. 1971 Civil disorder participation: a critical 7132012) relationship between personal relative deprivation
examination of recent research. Am. Sociol. Rev. 36, 12. Smith HJ, Pettigrew TF, Pippin GM, Bialosiewicz S. and aggression. J. Soc. Psychol. 157, 766–776.
1058–1073. (doi:10.2307/2093765) 2012 Relative deprivation: a theoretical and meta- (doi:10.1080/00224545.2017.1288078)
 17. Abbink K, Masclet D, Mirza D. 2018 Inequality and inter- 22. Kruschke JK. 2010 What to believe: Bayesian 2011 London riots. Eur. J. Soc. Psychol. 50, 10
group conflicts: experimental evidence. Soc. Choice methods for data analysis. Trends Cogn. Sci. 14, 646–661. (doi:10.1002/ejsp.2650)

royalsocietypublishing.org/journal/rspb
Welf. 50, 387–423. (doi:10.1007/s00355-017-1089-x) 293–300. (doi:10.1016/j.tics.2010.05.001) 28. Scacco A. 2009 Who riots? Explaining individual
18. Stott C, Ball R, Drury J, Neville F, Reicher S, 23. Sorensen T, Hohenstein S, Vasishth S. 2016 Bayesian participation in ethnic violence. Columbia University.
Boardman A, Choudhury S. 2018 The evolving linear mixed models using Stan: a tutorial for See https://search.proquest.com/openview/
normative dimensions of ‘riot’: towards an psychologists, linguists, and cognitive scientists. 7c75f21fd4ed9bb58991c034783c9713/1?pq-
elaborated social identity explanation. Eur. J. Soc. Quant. Methods Psychol. 12, 175–200. (doi:10. origsite=gscholar&cbl=18750.
Psychol. 48, 834–849. (doi:10.1002/ejsp.2376) 20982/tqmp.12.3.p175) 29. Wilkinson RG, Pickett KE. 2009 Income inequality
19. Reicher SD. 1984 The St. Pauls’ riot: an explanation 24. Davies TP, Fry HM, Wilson AG, Bishop SR. 2013 and social dysfunction. Annu. Rev. Sociol. 35,
of the limits of crowd action in terms of a social A mathematical model of the London riots and 493–511. (doi:10.1146/annurev-soc-070308-
identity model. Eur. J. Soc. Psychol. 14, 1–21. their policing. Sci. Rep. 3, 1303. (doi:10.1038/ 115926)
(doi:10.1002/ejsp.2420140102) srep01303) 30. R Core Team. 2019 A language and environment for
20. Smith ER, Conrey FR. 2007 Agent-based modeling: a 25. Epstein JM. 2002 Modeling civil violence: an agent- statistical computing. Vienna, Austria: R Foundation
new approach for theory building in social based computational approach. Proc. Natl Acad. Sci. for Statistical Computing.

Proc. R. Soc. B 288: 20203091

psychology. Pers. Soc. Psychol. Rev. 11, 87–104. USA 99, 7243–7250. (doi:10.1073/pnas.092080199) 31. Goodrich B, Gabry J, Ali I, Brilleman S. 2016
(doi:10.1177/1088868306294789) 26. McPhail C. 1994 The dark side of purpose: rstanarm: Bayesian applied regression modeling via
21. Aaldering H, Böhm R. 2019 Parochial versus individual and collective violence in riots. Sociol. Q. Stan. See https://mc-stan.org/rstanarm.
Downloaded from https://royalsocietypublishing.org/ on 28 January 2025

universal cooperation: introducing a novel economic 35, 1–32. (doi:10.1111/j.1533-8525.1994. 32. Makowski D. 2018 The psycho package: an efficient
game of within- and between-group interaction. tb00396.x) and publishing-oriented workflow for psychological
Soc. Psychol. Pers. Sci. 11, 36–45. (doi:10.1177/ 27. Drury J et al. 2020 A social identity model of riot science. J. Open Source Softw. 3, 470. (doi:10.21105/
19485506 19841627) diffusion: from injustice to empowerment in the joss.00470)