Reputation in Evolution Rosaria Conte LABSS/ISTC-CNR ICORE, Gargonza, Italy March 18-20, 2009
How explain cooperation if nonreciprocators are better-off within the group?
Solutions are usually found in social control. But why agents carry on and sustain its costs?
Previous theory (Conte and Paolucci, 2002) defined reputation as reported-on evaluation (human intelligence)
Here, speculations about how this theory of reputation can help in answering that question.
Strong Reciprocity (SR) as one example of social control. Open questions
Tasks in social control
Transmission of social evaluations and its impact on efficiency of social control
Social cognitive theory of reputation and expected impact of reputation transmission on social control
Examples of reputation transmission from ethnographic descriptions
Conclusions and questions still open.
Hypothesis and simulation evidence
Since 1999, Evolutionary GT (EGT) has put forward and tested by means of simulation the idea that the evolution of altruism was made possible in human societies (prehistoric hunterer-gatherers) by
Strong Reciprocators carry out and sustain the costs of punishing nonreciprocators.
From Bowles and Gintis, 2003)
Like theory of 2nd-order cooperation (Heckathorn, 1987; Oliver, 1993; Horne, 2007),
SR leaves some questions open:
By definition, costs of punishment < costs of cooperation.
Is this true? What about costs of retaliation?
Why agents other than victims carry on punishment?
Tasks in Social Control
Directly (memory of own experience or observation of others’)
Indirectly (transmission of evaluation)
Benefits and costs of identification
Direct acquisition of information
Benefits: moderate (insufficient info)
Costs: high (info paid at own expenses)
Others’ experience (observation; cf. Nowak and Sigmund, 1998 etc .)
Benefits: moderate ( wider info but still insufficient)
Indirect : transmission of information
Benefits: very high (much wider info, since one receives information in return)
Costs: moderate (lower than own exp., but higher than observation)
Communicative act , assumed to be moderately low and constant
Effect of information spreading to undesirable recipients ,
Potential retaliatory reaction to info transmission,
Benefits and costs of reactions
Defence/exclusion : avoid bad guys, exclude them from partner choice and communication.
avoid bad deals
prevent cheaters’ profits from such deals
Retaliation/punishment (for a distinction, see Andrighetto et al., in prep.): action aimed to damage nonreciprocators.
act of punishing
further reaction of recipient.
To sum up…
Reactions are as beneficial as costly.
Punishment is as efficacious as expensive,
Exclusion is less convincing, but self-protecting
Identification can be optimised by means of communication.
Spare costs of transmission to optimise social control…
So far, so good.
However, communication is not for free
spares the costs of acquiring information,
entails the costs of transmitting it
Is it possible to reduce such costs without reducing benefits?
Let see whether reputation, as a specific form of social intelligence, can help…
A Social Cognitive View of Reputation
Types of social evaluations Evaluations Image Reputation Set of evaluative beliefs about a given target Meta-beliefs about others’ evaluation
Social cognitive properties of reputation transmission
No personal commitment of speaker about nested beliefs’ truthvalue.
No responsability about their credibility and consequences (“I am told that…”)
Implicit source of rumour Indefinite author of evaluation
Effects on social control
Reporting on beliefs of indefinite source,
Prevents the target from providing discharge
Lower costs of reputation transmission for participants
it can be practised by anyone
prevents escalation of aggression and violence: people are likely to transmit reputation, because they hidethemselves behind indefinite source
Travels fast (badmouthing faster)
before the victim's innocence is proved, her reputation is spoiled.
Reputation Transmission in Traditional Societies: An Overview “ Anyone who has obeyed nature by transmitting a piece of gossip experiences the explosive relief that accompanies the satisfying of a primary need”.” Primo Levi "About Gossip," 1986).
Cooperation or competition?
Reputation transmission = gossip
With Gluckman’s study (1963), gossip became an object of study of its own in cultural anthropology
No consistent view (Levinson and Ember, Enc. Cult. Anth., 1997):
Social cohesion (group maintenance; Gluckman, 1963)
Social conflict (alliance against, Colson, 1949; etc.).
Shift of focus (Brenneis, 1987; Besnier, 1989; Brison, 1978; etc.),
the form of gossip
its features ,
the social characters and relationships involved, and
The interactive nature of Fatufatu among the Nakulaelae
Successful Nukulaelae gossips often “pause dramatically at strategic moments” (Besnier, 1989)
Waiting for interlocutors’ interjections or comments
on the scandalous content of narrative
Atoll in Pacific Ocean
Joint creation of Talanoa in Bhatgaon
Talanoa (tah-lah-NO-ah) is idle talk of Hindu inhabitants from the village of Bhatgaon in Fiji, as described by Donald Brenneis (1978).
Reputation is the kernel of social hierarchy.
Speakers and audience cannot be easily dfferentiated because gossip is created jointly by all participants
Social relationships in Talanoa
Gossiper towards target: limit perception of own identity
Gossiper towards gossipers:
don’t cause sanctions or retaliation
Let recipients form opinions “of their own”.
Third-person narrative in Talanoa : Bole
A requirement of talanoa emerging from transcriptions is the continuous and repeated use of the word “bole” (lit. the third sing. person of present tense of the verb “to speak”), used to
mean “I’ve heard saying” or “they say”
refer to an indefinite speaker or source.
In both cases the use of “bole” caused the speaker to keep a distance from what s/he says: s/he is not reporting on his/her own opinion but on voice or rumours.
Ambiguous and indefinite narratives in Talanoa
Talanoa transcriptions cannot be understood without previous knowledge
Heavy use of metaphors, irony, atc., communicating that what is hidden is more than what is said.
Targets are never clearly identified
The authorship of a particular gossip is blurred
Indefinite authorship among the Hopi
Bruce A. Cox (1970) studied gossip in the reservation Hopi in Arizona, 11 villages
Gossip starts when political authority is monopolised by one group.
Powerless use gossip to keep authority under control and form alliances.
Victims cannnot escape effect of accusation since,
source is not revealed,
evidence is not brought about
Hopi pueblo in Arizona (1879). Hopi House near Grand Canyon, 2005
“ Just talk” among the Kwanga
In 1992, Karen Brison studied the Kwanga, a tribe of hunterer-gatherers in Papua New Guinea which lives in numerous villages characterised by complex social networks.
Initiated men form a community of equals. Attempts to command lead to loss of support.
The Kwanga hold long community meetings to discuss matters of common concern), during which gossip is spread about powerful men:
If asked to produce evidence accusators resort to a conventional solution: they claim theirs was just talk , rumours…...
To sum up
Gossip leaves indefinite
The target ( Kwanga )
The source ( Bole )
Is spread in absence of target ( Talanoa )
Is unfalsifiable and unaccountable ( Hopi )
maintains group-values and identity ( Makah )
but also creates alliance (Hess and Hagen, 2002; Goodwin, 2002) of underprivileged against luckier
prevents retaliation (many)
it is fun ( Talanoa )
deplorable ( Fatufatu )
What these features amount to…
Impersonal narratives no commitment on truthvalue
Joint creation of gossip no responsibility
common in small acephalous communities (see also Boehm, 1999), where members are interdependent and act covertly instead of taking direct action which might offend others.
Ethnographic evidene matches the cognitive analysis
Any general conclusion?
Gossip is a universal behaviour, with more or less the same features, but different social consequences
Nonsense to look for consistent effects! These are often prosocial, but sometimes gossip may be used strategically, to create alliances against someone, etc.)
Its universal features converge on
transmit a reported-on evaluation (reputation)
Prevents retaliation thus sparing participants both
costs of info acquisition
costs of transmission
Provides incentive to participate in social control
Low cost transmission provides incentive to informational cooperation.
Hence, larger informational basin
What about material cooperation? No final answer. Under certain conditions, E.g.
Negative and unfalsifiable gossip + consequent exclusion:
Set of agents said to be bad implies and exceeds set of bad agents
The larger the informational network the lower the number of cheating deals .
To be checked by means of simulation!
Why inherently pleasant?
Perhaps because it is a self-enhancing protected aggression?
Motivations and emotions should be investigated…
unprivileged and weak use gossip as their only weapon
In natural reasoning, a material implication becomes an equivalence ( Geiss & Zwicky, 1971; Oaksford & Stenning, 1992; Wason & Johnson-Laird, 1972) . From, “if p then q”, to “If p then q, and if q then p.”
From “If you are weak, you can only aggress by means of gossip”, to “If you gossip, you are weak.”
Participants in gossip are said to be weak and unprivileged.
Hence, gossip is a vile and self-derogatory practice.
Dunbar (1998) and Panchanathan (2001) suggest that gossip evolved as an adaptive response to a selective pressure towards enlarging hominids’ settlements.
Here gossip is argued to have evolved
Taking advantage of a human cognitive capacity
By providing incentive to participate in social control because it lowers its costs keeping constant benefits (spreading meta-evaluation inhibits retaliation).
Data from traditional societies match this suggestion.
Enlargement of human societies as a possible side-effect.
To be done
If deceitful gossiper is not punished, to what extent does social control work? Trade-off between
Unfalsifiability of accusation
Utility of information
Possibly, there is a threshold above which informational cheating leads to system collapse.
An empirical question for future simulation-based studies.