Friedman, E. and Resnick, P. (2001). The Social Cost of Cheap Pseudonyms. Journal of Economics and Management Strategy, 10(2): 173-199

Discussion leader/summarizer: AycaObekci

Research Question:

1. What is the social cost of easy name changes in online communities?
2. What are the possible solutions to prevent the social cost of costless identifiers?

Key Points/Claims/Methods

Repetitive social interactions and reputation formation are two factors that enhance cooperation in communities. Knowing that the relations are long-lived and misbehaving can lead to negative consequences such as ‘revenge’ or a bad reputation, people tend to become trustworthy. However, it’s easy to change one’s name in online environments and to get rid of the consequences. Other members feel betrayed and even stop participation after misbehaviors; this is one cost of cheap identifiers. Suspicion to strangers is another cost. Members in online communities tend to distrust newcomers because they might be people who have misbehaved and changed their names. It would be more beneficial to trust people until they make the first malicious act, and enjoy the resulting rich social atmosphere. However, the paper proves in sections 2 and 3 that distrusting to newcomers enables more cooperation in equilibrium than initial trust.
(PaulResnick: this is a good summary; but the theorems also make a somewhat stronger claim: that distrusting newcomers enables as much cooperation as *any* other set of strategies that are in equilibrium. This was the part that was non-obvious and hard to prove, but which no one citing the paper seems to pick up on...sigh.)
(BenjaminChiao: For the stronger claim that Paul mentioned above, I see that as an application of the Coase Theorem. If we view the entitlement of initial trust as a form of property rights, the Coase theorem states that the assignment of right to which side does not matter in terms of economic efficiency (such as pareto or Kaldo-Hicks). Therefore, what appears to be the social cost is really not cost. He first asked the questions whether a polluting factory should be punished, or whether the herd raiser should be punished for damaging the neighbouring crop fields. He showed that whether the factory (or herd raiser) does or does not have the right to pollute (or damage), the wealth of the society does not change at all in equilibrium through a market. This is one of the most important results in economics and law this 100 years. The important point is that the right has to be assigned somewhere or equivalently whether we do or do not give the right to certain entity. The competition of any unassigment rights will lead to some effficiency loss.
Reference: Ronald Coase (1960), The Problem of Social Cost. J. of Law and Economics.
)

For proving that suspicion to strangers is more efficient for the society than initial trust, the authors build a model and use prisoner’s dilemma. The game is played repeatedly infinite times, with M players. The players’ payoff’s are displayed in a payoff matrix(p.178) In the basic model, if every player cooperates in every period, the amount of cooperation is 1, if every player defects in every period, the amount of cooperation is 0. If only fixed identifiers are used, the amount of cooperation is at equilibrium with cooperation of 1. If free identifier change is possible, this equilibrium can not be reached. In the case of malicious players or trembles (unintentional defects made by mistake), these players lose their identities at the end of the period and begin with a new identity. Two strategies are described: Punishment(PGTS or PFTS ‘public forgiving trigger strategy’) and PYD(Pay your dues). Punishment strategy tries to discourage malicious players from defecting, while PYD strategy rewards positive reputations rather than punishing negative reputations. Under PYD, when a newcomer plays against an old-timer, the newcomer plays C and the old-timer plays D; therefore the newcomer ends with a payoff of -1 and the newcomer with 2. The dues deter the old-timer from deviating, because he wouldn’t want to lose his identity and become a newcomer. In the model, “on average, veterans must receive expected payoffs sufficiently larger than the entrant’s payoffs to prevent someone from defecting and returning at the following period as an entrant” and “the most efficient way to create a differential between the value of being a veteran and that of being an entrant is by having a veteran defect against an entrant; since this transfers utility from the entrants to the veterans”

The defection against cooperative newcomers in the model is similar to shipment delays new buyers face at eBay, or humiliation or work freshmen face at fraternities. (PaulResnick: right!)

Possible solutions to prevent the social cost of cheap identifiers:

1. Disallowing anonymity: If people had to use their real identities online, they would try to become as trustworthy as possible to protect their real life reputation. However, anonymity is desired in online environments. Health support or political discussions, role playing games (where the point is to become another identity) and private online purchases are some situations where anonymity is essential even though it is accompanied with less accountability.
2. Use of entry fees: Entry fees would discourage people from changing identities often, however, there’s not the right amount of fee where rich people would find high enough while poorer people could afford. Entry fees also may keep people from joining the community in the first place, if they aren’t sure that the community is really beneficial and ‘just for them’.
3. Use of free but irreplaceable identifiers: In the system suggested by the authors, there is an intermediary institution that is responsible to provide members with once-in-a-lifetime identifiers. Public key encryption∞ is used to prevent the anonymity of the members (the intermediary can not know the real life identity of the person) while assuring that each person has only one identifier. (The intermediary checks the person’s real identity if it already has an online identity, and if so, doesn’t grant another) At the extreme case, there would be only one once-in-a-lifetime identity per person to use in the whole internet, but different identities for different social identities is suggested for it wouldn’t be desirable to use the same identity in different communities, say both in erotic literature and children’s literature forums.

Critique

<Are the major claims correct? Is anything important left out? Are the methods used to answer the question appropriate?>
Ok, with my limited math knowledge preventing me to fully understand the proofs, I rely upon the authors' verbal explanations and trust their math skills :)

One point about the suggested once-in-a-lifetime identifier system is not clear to me. I wonder:

• How will the collaboration among different online communities be obtained? How easily can different online communities coordinate to become a single "social arena", given that the communities or site owners might have different or conflicting interests, or their physical administration sites may be located in different countries. Maybe the site owners register their site as belonging to an arena at a .org site that serves this purpose?

(PaulResnick: there would be a registrar (a .org site) that you would get an identifier from. That identifier would be valid in only those sites that had signed up to be part of that "arena".)

Connections with other readings, ideas, etc.

• The quotation of Slashdot site creators in Lampe2004 (p.2) is a live example of the importance and desirability of anonymity in online communities. However, the site creators are also aware that anonymity may encourage misbehaving (low quality or off-topic comments in this case), and they have the anonymous contributors to “pay their dues” by assigning them lower initial score of 0 as opposed to standard 1.

• Dellarocas2004 discusses the implications of costless identities in section 3.4 (and cites this paper) He argues that with no cost of identity change, sellers will always cheat (and receive bad rating, and then change identity); buyers will expect them to cheat; and there will be no transaction at all. Therefore, in order to prevent the sellers to change identity and keep the market functioning, “new sellers should start the game with a state that has the lowest possible payoff among all states” [eBay doesn’t explicitly discriminate against new sellers as Slashdot does for anonymous users. I think the ‘state that has the lowest possible payoff among all states’ is not implemented as a design feature in eBay, but it corresponds to the effort and time required to build a good reputation and low profits expected as a new seller] Dellacoras also concludes that reputation mechanisms where players can costlessly change their identities should summarize fewer ratings to reach more efficient outcomes.

(PaulResnick: Right. If name changes truly were costless at eBay, and if every seller was strategic, we would expect a zero reputation to be the worst possible reputation you can have.)

• Dellarocas (2003) also refers to the social cost of cheap identifiers as discussed in this paper in a similar way. (Dellarocas C. 2003. The Digitization of Word of Mouth: Promise and Challenges of Online Feedback Mechanisms. Management Science/Vol. 49, No. 10, pp. 1407 – 1424)

• Novak, Raghavan and Tomkins. (2004) provide an interesting alternative solution to the problems created by changing identities. They study the identification of unique users among a set of online pseudonyms, based on content analysis; and report 87% success in matching authors with different aliases they use. If developed and widely used, this kind of detection systems might also be useful to detect and discourage the usage of multiple identifiers by the same person. (Novak, J., Raghavan P., and Tomkins, A., 2004. Anti Aliasing on the Web. WWW2004, May 17–22, 2004, New York, New York, USA. ACM 158113844X/04/0005.)

Discussion

LaurieBuis: Thanks to AycaObekci for providing a good summary... I found this reading to be quite challenging! Anyway, in addition to the critique that asked how communities will collaborate with one another, I have another to add... in the concluding remarks it is stated that "We expect both techniques for limiting name changes-entry fees and pseudonym commitments- to blossom in Internet arenas" (pg. 192). I find it difficult to understand how pseudonym commitments would ever be plausible in the Internet. Due to the nature of the beast, I just don't see how pseudonym commitments would ever gain popularity...