Whuffie

[gfish]

As I have ranted semi-incoherently about in the past, I'm a big fan of the idea of reputation economies. I find them more comfortable, less alienating, and compared with monetary systems the incentives are much better aligned with encouraging what I would like people to be doing -- being interesting and creative without being jerks. There aren't many environments in which reputation economies really work, but they're spreading, and I have high hopes for the future.

Naturally, I spend a lot of time thinking about the ramifications of how you implement such a thing in a large, heterogenous society. The best fictional example I know of is whuffie, which is the term I use for lack of anything better. The book doesn't really go into details, but it's an interesting structure. I'm not sure it would work properly with a simple mind-computer interface. Sure, being able to directly monitor someone's reactions would be a big part of it, but that only measures positive stimuli. I can give negative whuffie is someone is a jerk, or positive whuffie if someone invents something awesome, but I think it still fails the basic janitorial test -- do people get properly compensated for boring, unpleasant jobs like cleaning bathrooms? (Not that monetary solutions are particularly great in this domain!) I don't really notice clean bathrooms -- I just dislike dirty ones. In this case it isn't enough just to give negative whuffie to someone who didn't clean up after themselves. Having everyone scrub the toilet every time they use it would be very inefficient. Some things are best done in digest mode. And even if we were to do that, how much negative whuffie do they get? It would depend on the exact state of the bathroom before they used it, and how do you put an exact displeasure value on the incremental grungification of a bathroom after one use? It's silly.

The only way I can see around this problem is by having the brain implant be able to run counterfactuals when assigning whuffie. It isn't enough to ask how much I enjoy the clean bathroom, but somehow that has to be contrasted with how much I would have disliked it had it not been cleaned. It seems unlikely that this could be done in the concious brain, it would just be too distracting. So we need to be able to run simulations of your brain to see how they react to hypothetical simuli. This is starting to sound like the interface technology in Aristoi, where you could calve off background personalities to go deal with problems online.

The counterfactual assessments need to be time limited, however. It's unreasonable to keep accumulating whuffie for a bathroom that I cleaned a year ago. And at some point, the cognitive burden of millions upon millions of hypothetical scenarios is going to be unrealistic, even if they are being run offline in some subconcious corner of your exo-brain. If the bathroom is very dirty when I clean it, and then someone else tidies up slightly an hour later, they shouldn't reap my rewards, nor should that significantly diminish my whuffie. Now you need to run counterfactuals on if the second person hadn't tidied, if neither of us had done any cleaning at all, and on an entirely simulated scenario in which they tidied but I had never cleaned. I'm willing to grant the ability to do this to most people's satisfaction for a few layers, but no more than that. Hopefully this would only tend to happen in densely crowded areas, where the shear number of people appreciating your efforts would make up for the fact that they have to be dropped off the simulation horizon much sooner than normal.

Comments

[rollick.livejournal.com]

I honestly don't think the counterfactuals are the problem here, so much as the fact that the computer needs to keep track not merely of who dropped that toilet paper and didn't pick it up — so the janitor doesn't get the blame, since he did a good job cleaning before someone messed it up two minutes before you walked in — it also needs to keep track of which janitor cleaned the bathroom last, and thus who should get the positive whuffie when you think "Man, this place is impressively clean." Not to mention that it needs to figure out how to split the credit for the "This is a nice bathroom!" reaction between the janitor, the architect, the lighting designer, the contractor who actually built the place, and who knows who else. The only way I've ever been able to imagine this working is with a Central Computer out of John Varley's Steel Beach, something essentially omnipotent.

[caladri.livejournal.com]

I find that all of our existent reputation economies have serious problems with attribution. Just as a person with a lot of money can easily make more, those who have inflated reputations acquire credit and karma at an accelerated rate and often for things for which they have no rightful claim. As in rollick's point, the only solution is omnipotence or some set of rules that governs the distribution of attribution from a singular, dispassionate perspective (i.e. not from the perspective of the person who blames the waiter for their steak being undercooked, the person who blames the janitor for the filthy bathroom after a herd of Anarchists defiles it deliberately, etc.) At which point, why even defer to humans at all to determine who gets karma?

[stolen-tea.livejournal.com]

The sheer amount of processing power that this would require, boggles my mind. Hard enough to keep track of every human-generated event, and all of their real and potential consequences, but to measure all of this in some multi-dimensional karma scale?

I would posit that, should such a system exist, it would be compelling evidence for us being in the Matrix. :)

[gfish.livejournal.com]

The computers just need to keep track, the humans provide the karma values.

And, yes, stupidly complex. But I don't see any reason to think it's much more a couple of decades out of Moore's grasp. Tracking events won't take nearly that long. We're within 10 years of everyone constantly uploading video to a realtime StreetView/PhotoSynth system, after all. How long after that until someone throws together an app to AR-tag every piece of litter with the Facebook profile of the person who dropped it?

[gement.livejournal.com]

Where do you find existing reputation economies, aside from the obvious and somewhat problematic example of people becoming Internet-famous and voting each other's posts up and down? I want to poke at this idea more, but have little framework for it.

[gfish.livejournal.com]

They currently work best online, obviously. Any open source project. Wikipedia, or any of the hundreds of more specialized wikis out there. The OpenStreetMap project. But there are realworld examples too. Any volunteer effort is primarily motivated by reputation economics. And, of course, my adopted native city-state is built every year almost entirely that way.

All limited circumstances, but a lot more than there were 10 or certainly 20 years ago. We're still exploring the changing boundaries between what is best done monetarily and what works with reputation systems. I suspect there will always be monetary incentives, but I'm hoping we can replace large swaths with reputation as our tech (better communication, cheaper/easier fabrication tools, etc) continues to improve. Given a sufficiently post-scarcity technological base, I honestly think the world would come to resemble Burning Man.