LMSR subsidy parameter is the price of information

post by Abhimanyu Pallavi Sudhir (abhimanyu-pallavi-sudhir) · 2024-05-25T18:05:31.998Z · LW · GW · 0 comments

Contents

No comments

A logarithmic scoring rule to elicit a probability distribution $r$ on a random variable $X\in \{1\dots n\}$ is $s\left(r\right)=b\log \left(r_X\right)$. Something that always seemed clear to me but I haven’t seen explicitly written anywhere is that the parameter $b$ is just the price of information on $X$.

Firstly: for an agent with true belief $p$, the expected score from making a report $r$ is $E_p\left[s\right(r\left)\right]={\sum }_{x\in \{1\dots n\}}b\log \left(r_x\right)p_x=-bH(p,r)$ where $H$ is cross-entropy. This is maximized when $r=p$.

Well, this is just the standard proof that logarithmic scoring is proper. This max score itself is $E_p\left[s\right(p\left)\right]=-bH\left(p\right)$ i.e. the entropy in $p$. So your expected earning is exactly proportional to the information you have on $X$ (the negative of the entropy in your probability distribution for it), and the proportionality constant, the price of a bit of information on $X$, is $b$.

This can be made even clearer by considering the value of some other piece of information $Y$. If $Y=y$ and you learn this fact, you will bet $P(X|Y=y)$ which would give you an expected score of $E_{P(X\mid Y=y)}\left[P\right(X\mid Y=y\left)\right]=-bH\left(P\right(X\mid Y=y\left)\right)$. Taking the expectation over $Y$, your expected score if you acquire $Y$ is $-b{E}_{P\left(Y\right)}\left[H\right(P(X\mid Y=y)\left)\right]$ which is the conditional entropy $-bH(X\mid Y)$. Thus the expected profit from acquiring $Y$ is $-b\left(H\right(X\mid Y)-H(X\left)\right)=bI(X;Y)$.

So the value of $Y$ is precisely $b$ multiplied by its mutual information with $X$, i.e. $b$ is the price of one bit of information on $X$.

I assume this is widely known. But I think it’s still pedagogically useful to actually think in these terms because it sheds light on things like:

• Choosing a good scoring rule — it’s not just that we want the scoring rule to be proper (incentivize honesty), we also want it to incentivize the optimal amount of effort in acquiring information. $b$ should be a measure of how important a question is.
• But also: like the price of any good, the price of information can vary (e.g. you might want to reduce $b$ after some key information becomes public, since you’re getting it for free)! And like any good it would have diminishing returns. This motivates things like [1].
• It makes clear the fact that prediction markets have huge positive externalities — the market-maker is paying for the information, but it becomes public. This is bad (see also: [2]) — in general, IP rights remain an unsolved problem: [3]. I have a very clever idea to solve it, which I will elaborate in another post.

[1] “Market Making with Decreasing Utility for Information” by Miroslav Dudik et al. https://arxiv.org/abs/1407.8161v1

[2] “Transaction costs: are they just costs?” by Yoram Barzel. http://www.jstor.org/stable/40750776

[3] “IP+ like barbed wire?” by Robin Hanson. https://www.overcomingbias.com/p/ip-like-barbed-wirehtml)

0 comments

Comments sorted by top scores.