Tuesday, November 22, 2011

Reputation


Draft. Do not quote without permission. Submitted to the SAGE Encyclopedia of Philosophy of Social Science.

Reputation, from the verb puto in latin, meaning “counting, considering” plus the suffix re- that indicates the repetition, is the consideration of the value of an agent by other agents based on his or her past actions and creating expectations on the future conduct of that agent. Reputation is a special kind of social information: it is social information about the value of people, systems and processes that release information. Reputation is the informational trace of our past actions: it is the credibility that an agent or an item earns through repeated interactions. If interactions are repeated, reputation may conventionalize in “seals of approval” or disapproval or social stigmas.

The notion of reputation in social sciences has been mainly treated in economics. In Adam Smith’s liberal social theory, reputation is seen as a way of coordinating activities in a decentralized social space of transactions. According to Smith, in a free society, markets coordinate diffused knowledge in an asymmetrical way: people have a partial view of what other people know and how they will act. Also, given that most transactions occur over a lapse of time, parties have to trust each other that they will satisfy their reciprocal interest. These informational and temporal asymmetries call for efficient means of storing and retrieving information about possible partners in interactions. Reputation is more than pure information: it is evaluated information, that is, a shortcut of the many judgements and interpretations that people have cumulated about an actor. That is why people are interested in keeping a “good” reputation by signaling to potential business partners their trustworthiness.

In the rational choice tradition, reputation is modelled as a repeated game. These games raise the question on how you can signal your reputation before any interaction. That is, how you can signal your credibility in absence of information about your past behaviour. This question is studied within a rich body of work that goes under the name of Signaling Theory (Gambetta, 2009). Signaling Theory aims at solving a fundamental communication problem: Given an interaction in which interests diverge between the two parties, how can a party be certain of the qualities of the other party? Honest signallers will try to signal their good qualities (trustworthiness, accountability, strength), but dishonest signallers will try to do the same, by mimicking high-quality signals. Signaling theory may be traced back to the work of the American sociologist Thorstein Veblen. In his Theory of the Leisure Class, published in 1899, Veblen explains the display of wealth of the leisure class (luxury, expensive clothes, time-consuming unproductive activities such as sports) as a way of signaling its social position. Important developments of Signaling Theory go from the study of behavioural ecology (Krebs and Davies 1998) to the sociology and the economy of cultural tastes and lifestyles (Bourdieu, 1984). An agent emits signals in order to make a threat or a promise credible. Costly signals and robust signals, that is, signals that are difficult to fake, are those considered more credible (Zahavi, 1998).

The economist George A. Akelrof has shown that quality uncertainty is such a risky feature of markets, that reputation is needed: “Seals of reputation” in a markets are labels, certifications, guides, that is, all the devices that tend to reduce the informational asymmetry. A rational agent, according to Akelrof, has an interest in embodying these devices in order to compensate the cognitive deficit of the informational asymmetry.

Quality uncertainty and informational asymmetries have become crucial epistemological issues in contemporary information-dense societies. The vast amount of information available on Internet and on the media makes the problem of reliability and credibility of information a central issue in the management of knowledge. Informational items that do not come with some label, or seal of approval from the appropriate communities, are lost in the data deluge of the Information Age.

From the evaluator’s perspective, that is, the agent who has to filter information, reputation has an informational value. This has become a prominent issue in Web studies. Given that the structure of the Web is that of a reputational network, in which each link from a page to another can be read as a “vote” from a page to another, a number of algorithmic techniques have been developed to compute reputation of different entities on the Web: Recommender Systems, Collaborative Filtering and Reputation Systems (Resnick, 2000).

Collaborative forms of sharing ratings are also relevant in the study of Collective Intelligence (Landermore & Elster, 2012). People do not share information: they share evaluated and classified information that creates a “reputational stream” of shared judgements. The epistemological implications of the massive use of shared ratings in networked societies are huge: relying on other people’s judgements and authority challenges our epistemic responsibility. The reasons we trust collectively filtered ratings about an item or an agent are seldom explored. Choosing a doctor, an academic institution or a wine is a way of endorsing a tradition of values, a way of filtering information that is not always transparent and legitimate. Notorious biases in social networks - such as the Matthew effect, investigated by the sociologist of knowledge Robert Merton, according to which the nodes of a network that are more prominent have more probabilities to earn more reputation - create noise in the way reputation is diffused.

Other biases need further epistemological and cognitive inquiry. For example, people tend to form beliefs in order to acknowledge previously established reputations, such as voting for a certain party because a very well-reputed friend votes for that party. Also, reputations are resilient and may last over time even when the facts of the matters they are supposed to signal are no more there. For example, the prestige of institutions and corporations may last long time after their decay.

Reputation is a social commodity that needs to be handled in scientific way in order to avoid informational cascades, conformism and the perpetuation of received views.

Further readings:

Akelrof, G. (1970) "The Market for Lemons". Quarterly Journal of Economics, 84 (3), 488-500.

Bourdieu, P. (1984). Distinction. A Social Critique of the Judgement of Taste. Routledge and Kegan Paul.

Gambetta, D. (2009). "Signaling". In: P. Hedström, P. Bearman (eds) Oxford Handbook of Analytical Sociology, ch. 8.

Krebs, J. R., and Davies, N. B. (1998). An Introduction to Behavioural Ecology. Blackwell, Oxford.

Origgi, G. (2012)." Designing Wisdom Through the Web. Reputation and the Passion of Ranking". In H. Landermore, J. Elster (eds) Collective Wisdom. Cambridge University Press.

Veblen, T. (1899/2007). The Theory of the Leisure Class. Oxford World’s Classics.

Zahavi, A. (1998). The Handicap Principle. Oxford University Press.