6 ms·
Of course, but that's the whole point. Rather than debating policies which are nominally designed for one goal but secretly trying to promote another goal (e.g.
by yummyfajitas 13y ago
Of course, but that's the whole point. Rather than debating policies which are nominally designed for one goal but secretly trying to promote another goal (e.g., social security, nominally a pension, secretly a redistribution scheme), we just debate the goals themselves.
This actually increases transparency. For example, if some party wants to declare a War on Christmas (TM) or a War on Women's Bodies (R), they actually need to put 23 x Dead Santa or -6 x Women's Freedom into the objective function.
- notahacker 13y agoI think people are pretty concerned about means as well as ends. That's why all parties want to reduce crime, and yet none campaign for a policy of "dispose of crime reports after 24 hours" which is almost certain to fulfil the objective function for "reduce recorded crime" (and will probably get passed in "futarchy" if someone proposes it and the market acknowledges that the highly-weighted variable "crime rate" will undoubtedly fall as a result). Instead of voting for the party that wants to reverse something that directly impacts their welfare or they otherwise consider to be stupid or immoral, the electorate is forced to listen to a set of explanations of mathematical equations that try to reverse engineer how it could have been passed, and how to remove it with the least side effects. If I want to pay lower taxes I'd prefer to vote for the party that makes lower taxes for people like me a core plank of their platform, and not a set of equations purporting to show how a tax cut could feasibly be approved by the market in the current economic climate provided variables P, Q, R and S were held constant and Z was reduced by a couple of basis points. It's quite possible that a party advocating a high weight on lower-quartile average incomes and a negative weight on birth rate wouldn't want some nutters to propose the "compulsory abortions for offspring of the unemployed" law, and get it passed, but I'm not sure a welfare function that could avoid all such unintended consequences even exists, never mind one intelligible to the average voter.
- RobinHanson 13y agoStatistics about means can also be included in welfare measures.
- notahacker 13y agoYour article explicitly argues against that Courts should invalidate any welfare definition change that seems too directly an attempt to favor a specific policy whose main rationale comes from other measurable outcomes That - and your "miles of road" example - seems to me to invalidate changing the welfare definition to add welfare weightings for tax levels.
- RobinHanson 13y agoPutting in weights to directly favor specific policy proposals is very different from putting in weights to penalize the kinds of means that people dislike. For example, if we dislike police detaining folks without charging them with a crime, we could count such events and penalize them.
- yummyfajitas 13y agoThe objective function would need to include the measurement methodology, i.e. "reduce crime as measured by methodology X", where methodology X does not delete crime reports. As for voting for tax cuts, you probably want taxes cut as low as possible while still providing certain vital services. (I'm assuming you aren't an anarchist.) In that case, you don't merely want maximize (-1 x tax rate), you instead want to maximize (-A x tax rate - B x # of murders). Lowering taxes may or may not be the best way to achieve that goal, but you are unlikely to be able to properly judge that (no offense intended - I can't either). An objective function and it's unintended consequences may be hard to understand for the average voter but the results of policy are even harder to understand. For example, many voters are in favor of earning assistance for firefighters and soldiers (http://www.opencongress.org/bill/110-h3997/show http://www.opencongress.org/bill/110-h3997/show). One unintended consequence of the "Earnings Assistance for Heroes" policy is a gigantic bank bailout.
- notahacker 13y ago>An objective function and it's unintended consequences may be hard to understand for the average voter but the results of policy are even harder to understand. Reverse that statement and you've got my position. It's very difficult for an average voter to figure out where the cash for his tax cut is coming from, but it's crystal clear approximately how much disposable income they will personally benefit from it even if the indirect consequences are pretty hazy. It's up to the government to figure out how to pay for it, and to balance that out against the desires of people that are really far more concerned about murder rates, welfare cuts or national debt rises. By contrast this system asks voters to become technocrats and carefully predict the set of variable weightings that will allow the investors to pass legislation that reduces their taxes; that is if Big Capital doesn't decide to bet on a policy proposal that uses the freed-up budget to reduce their taxes instead. Frankly, if an average voter can't determine if they want the candidate that offers them lower taxes instead of the one that offers them better policing, they're certainly not capable of choosing between the two candidates' proposed set of HDI weightings and calculation methodologies. Nor does anything in the proposed mechanism stop policy proponents sneakily attaching undesirable policies to desirable ones: in this case if you offered the banking system bailouts they'd probably bet big enough on the bill to allow it to pass even if it had subsidiary clauses mandating the execution of firefighters' firstborn. It would be trivially easy to use this process to enact laws whose consequences were essentially non-statistical, like most civil liberties...
- DennisP 13y agoHm. I'm warming up to it. A remaining problem might be actually measuring those values in ways that can't be gamed too easily. Attempts to objectively measure programmer productivity have generally failed at this, and Women's Freedom seems harder to measure than that. Another problem might be that the objective function will keep changing. I guess the way to handle this is to pay off bets based on what the function was when the bet was made, and to have a fixed schedule that alternates making bets and changing the function.