bowaggoner

I enjoyed the comment. You might like my short story along the same lines from a few years ago: https://bowaggoner.com/writeups/robust.html

This project is more on the academic side -- disclaimer, I was involved in it, but it's led by Justin Harris at Microsoft Research: "Sharing Updatable Models on Blockchain" https://github.com/microsoft/0xDeCA10B

The idea is that the smart contract is a learning algorithm, and people can donate data to a public repository stored on a blockchain. The learned model is publicly available for everyone. People can also receive incentives for donating data in some implementations.

I respect that perspective, but did you know that Signal uses a private protocol so that your contacts are never sent to their servers?

https://signal.org/blog/private-contact-discovery/

That's one big problem for sure. Another is when the information is costly to gather, and there is a Schelling point for people without the information. For example, we all get in the habit of saying "10 degrees" whenever anyone asks the temperature of any city.

Similar, there are cases like "What is the capital of the state of New York?" A Schelling-point answer might easily be New York City, even though the correct answer is Albany. There's research on this stuff in the "peer prediction" a.k.a. "information elicitation without verification" literature, e.g. Prelec, Seung, McCoy 2017: https://marketing.wharton.upenn.edu/wp-content/uploads/2017/...

Prediction markets also exist for non-binary events, whether a finite set of outcomes, countably infinite, continuous, or beyond.

Instead, the problem is they still require an oracle for the actual event outcome. That is, everyone is betting on some event, but once it happens, an oracle is needed to tell the market what the actual outcome was. So they don't solve the oracle problem. They're another application that can make use of oracles.

Should have asked for a dot "somewhere" on the page.

Oops, thanks!

Thanks, added!

Hi all, seems like a slow news day so thought you might enjoy this little page. Comments, suggestions, discussion, criticism welcome!

Great point! We ultimately felt (and this is above my paygrade) that it was outside the scope of what we were asked to explore/recommend. So we avoided suggesting taxes that drastically change corporate incentives and behavior around private data. Instead we focused on more equitably distributing the value that's currently generated from data.

We did discuss this and felt that a very deep dive would be needed. (A) It's not an objective no-brainer exactly what behavior you want to disincentivize. (B) The economic impacts/consequences could be huge, so the tradeoff needs to be carefully considered.

One example is Google Maps. Right now it has privacy drawbacks, but it also generates a lot of utility for a lot of people and is relatively freely accessible for individuals. You'd want to be careful about screwing that up. For example, incentivizing Google to switch to a paid access model.

I guess part of what this EFF article is saying is: these decisions are better treated as part of privacy law / human rights, rather than economics...I'd be interested to hear your thoughts!

Thanks for the reply - interesting concern. By analogy, I don't think putting taxes on cigarettes means the state will try to get more people to smoke. I also think it depends on how the money is allocated. Your concern would be strongest, I think, if the money went into some general-purpose fund. But if it funds open-source projects or a data-relations board, then I don't see much cause for concern.

Very nice article!

Last year I was part of a working group on a data dividend proposal for the state of California, which you can read about at https://www.datadividends.org/ .

Our conclusions led us much closer to the spirit of this EFF article than the term "data dividends" might suggest. We recommended against any kind of personalized payments. We considered a small universal basic income funded by a "data tax", but because of the small amount (as mentioned by the EFF), we focused on use of a data tax to fund public projects and initiatives aimed at redistributing the benefits of data more equitably.

We mostly stayed away from recommending privacy or data ownership regulation because (a) California had recently passed the CCPA and (b) this question seemed outside of our mandate, but I agree that the two should be considered together.

Hi all, we are a couple CS researchers and made this game purely for people to have fun.

It's designed to be accessible to anyone who has never heard of a prediction market. In fact, I would love to hear if you think this is a good prediction-market tutorial.

(Submitter) Hi all, my friend and I are researchers in game theory and computer science.

We wanted to make a site where people could have fun predicting Game of Thrones outcomes, without needing to know at first what a prediction market is.

I'm curious if this can be a good way to teach people about prediction markets. And most of all, we want people to have fun!

Any thoughts, suggestions, etc. appreciated!

Unfortunately no video, sorry. Glad you find it interesting!

You're right and we're on the same page about the solution. The reasoning behind my comment of "interesting" is that formula you gave is not well-defined for p=infinity, although we can define it as the limit of that expression as p-->infinity. Furthermore, for p < infinity the lp-norm can assign a well-defined penalty to each x, and the goal is to minimize the sum of the penalties. But that's not really true for p=infinity.

Unfortunately MathJax is a bit large and nontrivial to host yourself. And it recently switched to being hosted on cloudfare instead of mathjax.org.

I struggled with this choice myself and so far, decided to do the same as the author with a noscript tag to explain to the reader why I'm loading cloudfare code.

Hmm, that example is really interesting. I think you're right once we formalize L_infinity, because the goal will be to minimize the maximum possible distance. But the other scores can be phrased as penalties and minimizing expected penalty or total penalty summed over the data points, and it's not clear how to phrase L_infinity this way because the "penalty" would be infinity for any outcome...

There is a much wider generalization here which is studied under the name of "property elicitation" in computer science, machine learning, and statistics.

The generic question is: Given a loss function, what "property" of the distribution minimizes average loss; and given a "property", characterize all such loss functions. For example, Bregman divergences are (essentially) all losses that "elicit" the mean of a distribution. If you have any monotone continuous function g(), then |g(x) - g(s)| actually also elicits the median, and these are essentially all that do.

Apologies for self-promotion, but you can read more at references on this page (disclaimer: I'm one of the researchers who posted it): https://sites.google.com/site/informationelicitation/

or tutorials on this subject at my blog: http://bowaggoner.com/blog/series.html#convexity-elicitation

Both terms are fluid. Most theoretical computer science research is essentially mathematics, at a high enough level. Some wouldn't necessarily call themselves mathematicians unless their graduate training was in a math department; some might. Some would call themselves scientists insofar as they consider mathematicians a subset of scientists. Most would not consider themselves scientists in the sense of natural sciences like biology.

But, most probably don't have a strong opinion on whether/which labels apply! They would say they just do research in TCS.

CS as practiced today contains elements of math, engineering, and the sciences (though rarely at the same time). I have a bit on this you might find interesting: http://bowaggoner.com/whatiscomputerscience.html#medium

Thanks - I took out a bunch of extraneous transition slides (except in one place where I think it's helpful to see).

I recently was researching bitcoin, blockchain, etc, and I found it surprisingly difficult to find good but relatively short resources! I synthesized what I found into some slides that you might find helpful[1].

Apologies if this seems like self-promoting my own slides -- and I should say I am not at all an expert here -- but I created them mainly because I had trouble finding a "primer", so maybe they can help others in a similar position.

[1] http://bowaggoner.com/talks/2017/bitcoin-etc.pdf