A recent post by Bitarch on trust networks reminded me of my decentralization cycle, and that it would be a good idea to supplement it with a new note.
People inevitably interact with one another and would like to derive maximum benefit from this with minimum costs. Of course, in a praxeological sense—meaning it doesn’t necessarily have to be about monetary profit or loss, but rather about satisfaction or, conversely, dissatisfaction. In repeated interactions, one can attempt to predict the counterpart’s actions and, based on this, decide in advance how to act. And if the counterpart knows that the interaction will continue to repeat, this knowledge will also influence their actions. Constantly interacting counterparts can agree on cooperation, distance themselves if they find each other unpleasant, send signals suggesting the boundaries of what is acceptable, and so on. Each person inevitably builds up a certain reputation in the eyes of the other.
Additionally, in human communities, the transmission of information about one person’s reputation to another person is of great importance. The other is not obliged to take this information on faith and may, on the contrary, believe that they are being misinformed for some reason. And even if the information received turns out to be completely reliable, they may still believe that someone is trying to gain their trust only to deceive them later. There can be as many layers of reflection as one likes; that is why we grew such a complex brain—to gossip with pleasure.
To gossip and especially to collect gossip about people outside their constant circle of communication, people devise various systems for recording and representing reputation. What should such a system be like from the perspective of someone studying another’s reputation? They would want to be able to receive information about a person’s behavior in the area of interest, and for that information to be easily accessible, difficult to forge, and difficult to hide. And if it is not a matter of studying one specific individual, but of searching for a suitable option, then there should be many options—ideally all possible ones—with convenient data filtering tools.
And what should the system be like from the perspective of the person whose behavioral data is being studied? Naturally, they would prefer that only the information they want to keep public is made public, while information that doesn’t need to be known by everyone could ideally be formatted as reliable proofs and presented only to those who need it. It would also be useful to have the ability to provide ill-wishers with convincing fake information about oneself.
From this set of contradictory requests, it is clear that an ideal reputation system that satisfies everyone is impossible. Again, we have so far neglected the costs of collecting, storing, and presenting reputational information, but those costs also exist.
Thus, it is quite easy to imagine the economically justified construction of a reputation system in some specific narrow field, but it is difficult to understand who would be interested, and why, in maintaining comprehensive open dossiers on every action of every person. Even government agencies prefer to collect voluminous dossiers only for official access, and their employees are certainly not interested in ending up in them themselves, perfectly understanding that the wider the circle of those who have voluminous information about them, the more vectors there are for sudden attacks.
Who, besides the consumers of reputation information, is interested in this information appearing in the corresponding systems? Of course, those who believe that their reputation in this system will present them in the light they wish to be presented. Those aspiring to the title of faithful spouses are unlikely to want to see themselves in an open registry of top lovers, although in another context, a high rating in that capacity would flatter them. A talented craftsman will have nothing against appearing in a rating of masters in the relevant field, while a less talented one would prefer to appear simply on a bulletin board, without a built-in reputation system, or at least have the ability to boost their rating a bit.
What general conclusion can I draw from all these considerations regarding the processes of centralization and decentralization in reputation systems? In a free market, it seems more likely to me that there will be many local reputation systems, rather than comprehensive, monstrous social rating systems. Even if some open system stores data on a large number of people, this will be fairly narrow information. Services for aggregating comprehensive open information on a specific person of interest are also quite monetizable; however, the objects of such research are unlikely to be interested in having information about them from diverse sources be easily aggregated. Nevertheless, working with big data today is a relatively inexpensive pastime, so few will be able to seriously confuse researchers of someone else’s digital footprint.
But these numerous local reputation systems will, I believe, be mostly centralized. As for systems of distributed storage of reputational information in electronic form, they are unlikely to gain serious popularity, remaining a niche solution for small communities. There are no serious prerequisites for scaling here.
