Lift11: Azeem Azhar, Online communities and reputation management [en]

[fr] Notes de la conférence Lift11 à Genève.

Live and India-lagged notes from the Lift11 Conference in Geneva. Might contain errors and personal opinions. Use the comments if you spot nasty errors.

Reputation and its importance for communities. Azeem has known Chris for 11 years on a mailing-list, and this is the first time they’ve met in person.

The reputation graph. Plays a very important role in helping markets function.

History: initially we all came from very small communities. There was no trust/reputation problem, we knew everybody. No issue. If you didn’t know somebody, chances are they were a threat. In that world, we could hold all the relevant and necessary information in our heads. (cf. Dunbar)

The world has changed for most of us. Hyperconnected world. Facilities to connect to almost anybody we want to.

The cost of connections and making connections has really drops. So we have lots of connexions to a lot of people that we don’t really know. Azeem has 130 friend requests on Facebook from people he doesn’t know.

We’re pushing past the barrier where maybe it’s not that useful to have so many connexions, because the trust element diminishes.

Stock market: lots of people who don’t know each other very well transacting with each other. How does that work well? (1) Regulation (highly regulated! penalties for people who break the rules). (2) Contractual: standards bodies saying what things must mean when you say them. (3) Reputation rating: looking at activity in a firm and coming out with a rating from AAA to BBB or below. This lets a participant look at a complex company and evaluate the risk of making a trade.

What happens when people stop believing in ratings? that’s what happened in 2007-8 in financial institutions. “JP Morgan, I don’t care what your rating is, I’m not lending you money” => liquidity dried up, huge crisis requiring government to step in. This gave us an experiment to see what happens in a market where reputation systems break down. Answer: it’s really really horrible.

Other context: chess world, rating players. Determines what kind of competition you can enter, and what you stand to win or lose in any given match.

Academia: ratings help you win grants or attract talent.

When I buy a coffee at Starbucks, I can’t trust the person at the counter because the person is a good person, but because of the trust and reputation that Starbucks has built over the years. But we’ve learnt that this kind of reputation is built around other objectives and can break down easily.

Trust ratings in the web, Google. PageRank *steph-note: breakdown there I’m suffering from right now with CTTS, pages being unindexed because my site was hacked a few months ago and pages were full of spam*

But now the web is about people connecting to each other, not just about pages. What we need is a PeopleRank, to make sense of the connexions between people. Cf. Quora. We can think about doing this because now we all live in public. Huge set of data to make sense of.

Foursquare: move towards rewarding people on platforms by giving them badges, but it rewards activities and hoop-jumping — artificial activity. Reinforces the business model of the platforms (use the site more and more and more) but doesn’t really tell you anything about how trustworthy people are or if they really know anything about art galleries because they have the art gallery badge.

eBay: nice but not portable. Very context-dependent.

Other approach: get your friends to recommend you (LinkedIn) — but friends can be bribed. Recommendations cluster around the time people leave their job.

Azeem thinks we’re moving towards a single currency. Portable, handles different contexts, and has value. It’s what happened in the corporate bond market.

Peerindex. *steph-note: mine is 40 seems very limited and skewed towards Facebook/Twitter/LinkedIn — doesn’t take into account my blog where most of my reputation is built, and I for example neglect linkedin recommendations completely*

What are the bad applications of reputation? Health insurance discrimination.

We get a fantastic benefit from Facebook etc, but once the reputation stuff starts working, a lot will start happening in matter of transactions.

Bond ratings: a single number, but tries to predict one thing — how likely is the company to defect on its bond payments?

We’re starting to feel a move towards quality of contacts rather than just hoarding them. Path for example, which allows you only 50 friends.

Popularity does not equal trust. But there is maths that allows to factor popularity out.

Similar Posts:

Leave a Reply

Your email address will not be published. Required fields are marked *