Goodhart's Law - Substitute Goals

Philosophy Forums Society, Government, and Economics
1

There’s an old story, not sure if it’s fiction or fact, about Communist Russia:

They wanted society’s nail producing factories to be efficient, so they set a goal to measure the efficiency of nail-producing factories. Problems were rapidly apparent:
When efficiency was measured by number of nails produced, the factories produced millions of unusably small nails.
When efficiency was measured by total weight of nails produced, factories produced a few incredibly large nails.
When efficiency was measured by value of nails produced, factories produced golden nails.

So, what’s happening here is that there’s a goal, G, that is trying to be met (presumably something like ‘efficiently produce nails that people want’), but instead of setting G as the standard (because maybe G itself is hard to measure, or something like that), they choose a measure that was correlated with G, G*. So, prior to Communism, presumably highly efficient nail factories COULD be measured based on number of nails produced, weight, etc. Because, prior to communism, those things weren’t the goals – the goal was to produce nails that people wanted, so that you could sell them for a profit.

Take profit out of the picture, and ‘producing nails that people want’ suddenly doesn’t look so important. If the government says ‘produce as many nails as you can’, you make as many nails as you can, without concern for if people want them. If the government says ‘produce as much weight in nails as you can’, you produce as much weight in nails as you can, without concern for if people want them.

This is Goodhart’s law in action. Goodhart’s law has a number of formulations:
“When a measure becomes a target, it ceases to be a good measure.”
“As soon as the government attempts to regulate any particular set of financial assets, these become unreliable as indicators of economic trends.”
“Once a social or economic measure is turned into a target for policy, it will lose any information content that had qualified it to play such a role in the first place.”
“Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.”
“When a measure becomes a target, it ceases to be a good measure.”

Or, in other words, if some measure G* is highly correlated with some desirable goal G, and so you start rewarding G* in an effort to maximize G, the correlation between G* and G rapidly deteriorates.

2

An interesting example is in evolution as it relates to Humans:

We have evolved rewards to having sexual intercourse – pleasure, orgasm, etc. Sexual intercourse is G*. Producing surviving offspring is, as far as evolution is concerned, G, right? So, evolutionarily speaking, if a species has a reward for sexual intercourse, which tends to produce more surviving offspring than the same species without the same reward (in other words, a species that has no pleasure from sex will reproduce less presumably, because there is less incentive to have sex), then evolution favors that reward.

And so we, as humans, know that having sex is rewarded, and so we as humans do it often. To achieve that reward, right? But sexual intercourse in humans has ceased to be correlated with reproduction – we’ve hacked the system. We have condoms and pills and vasectomies, etc. So we can and do achieve G* – sexual intercourse – and bypass G altogether. The relationship between G* and G, then, ceased to be because G* was rewarded independently of G.

3

So that’s the ‘G’ Spot then? :slight_smile:
Sorry, I couldn’t help myself. You may delete my fickle response.

4

I am on a role with deleting everything in a post before posting it by hitting CTRL X instead of C when I try to save it prior to posting.

Okay, very fast:

In antiquity, there was a massive divide in number theory… many claimed 1 and 2 were not real numbers, that the first real number is 3. It’s part of the basic idea your presenting… 1 is potentially nothing- it’s a reference point PERHAPS, or even a action. A self referential metric doesn’t even begin until number 3, and that’s when other cognitive capacities kick in. A reference point can be nearly null of attributes, but more complex numbers have to be counted, and have greater amount of relative reality associated with them beyond a linear quatity of type or category… two nails might be bent for example, or shiny, or dull, or they might be potato chips instead of Nails, because Vasili down the line likes hit potatochips on the job, and they knocked the nails that were suppose to be there off the assembly belt.

This is some very basic, platonic era dialectics. Civilization adopted the weaker system, and as a result, we’re less aware of the inherent discrepancies in perspective and thought… we can choose to assume one psychological function is enough. Worst, in revolutionary group think, the situation can become a self congratulating, yes man, circle jerk heading nowhere. The people attracted this this outlook aren’t too different from Sauwelios basic typology (not his philosophy he advocated, but how he advocates it- in a dry, uninspiring, and essentially broken and dead outlook)… he gets excited at the technical understanding of the linguistics underlining utilitarian thought (his greatest issues are in circumstances when men universal advocate a WTP by everyone as non-frictive to each individual’s will, he picks up at least this is not the case, but it’s the deepest I’ve ever seen Sauwelios in terms of active contemplation- it’s a utilitarian function of the mind in the left hemisphere)

You get these people together into groups, such as a revolutionary council or a people’s committee (or DHHR in the US with governor’s 5 year stategic plans and the need for census and concensus) and they are essentially going to be blind to the attribution of the materials in relationship to technical thought. They know technical thought exists, and they are aware- from bullshit emperical data as well as ideological assumption, that they want to maximize the best positive ideal possible in material terms… and that it can be done through concensus and reason, and that the byproduct is needed by the consumer.

Problem is- such a bottom up approach to such a empirical outlook is incapable of accurately guaging what the ‘consumer wants’, or even if they want to consume this crap being pumped out. The bottom line is abstract, and the whole system becomes abstract.

The soviets tried to fix this in the 1930s with two branches of literature:
Social Realism: en.wikipedia.org/wiki/Socialist_realism
they also introduced a laughable genra of heroic literature of worker innovation who fight the system to introduce reforms of manufacturing, such as a new kind of release valve, despite the manager’s resistance and workers chiding them.

In the first, social realism- it’s similar to the roaming dialectic of the Cynic- the confrontation of the real produces approtunities to understand themselves and the world they live in under new circumstances… but given they were hegelian ideologically, the soviets made certain the path the dialectic took them was to a higher awareness of socialism and class consciousness. You can see the roots of this in Jack London’s Socialist Writtings or even Dickens… but it wasn’t up to the challenge of making quality reflect in quantity, as a reference to market demands were arbitrarily established and usually without much foundation.

In the second… most everyone gave up. The soviets got around this issue in later years by brain farming… they would put their engineers in isolated communities with more luxury goods, and it proved to be liberal and rational enough to develop new cutting edge technology of multiple and diverse forms. However, they never really made stuff for useful, non-military applications- you keep a population isolated, they are not in contact with the real world… and they got to go off assumptions and memories. Besides, the budgets they wanted was for the big ticketed weapon systems, not for redesigning a more useful radio.

Hence why it failed. Perestroika couldn’t resolve these inherent issues. However, basic engineers in the soviet union were good at reverse engineering technology, and they did advance somewhat… though it was always sluggish in regards to the west.

What’s interesting is how a similar educational institution- the university system, is creating parallels in the West as in the old Soviet Union. Many of our top thinkers are isolated from the population, in gated communities. The rise of Obama style socialism has caused a even more rapid decline in workers feeling affiliated with their work place, less incline to pitch in or suggest ideas they are not required to given worker loyalty is so low now. It’s hard for someone like me who’s background in management is Servant Leadership to look at the current situation economically and administratively as positive. We’re essentially declining into a soviet style government with the bail outs coupled with presidential mandates for new kinds of technology- the Tesla Electric car is a good example. Considerably more complex than a nail, but within the same paradox. Meanwhile, we can import electric bongo trucks from Japan and use them here just fine under industrial/commercial conditions.

The utilitarian mode of thought, in it’s summation, is ineffective and unwise. It requires other forms of thought to make it useable. We’re were once culturally conditioned to know this, but as time went by, and utilitarian thought flooded out univerities… we’ve intellectually devolved. It’s a relatively small portion of the population working on designing computers and smart phones- and most everything else we have has been progressively turning to shit. Since classic times, such as with our current number system we unfortunately adopted, we’re less aware of the inherent issue lying at the heart of the problem. We can’t see the problem- we see other kinds of problems as a result, and can only come up with half assed solutions.

The good news is, a society can potentially pull their head out of their ass whenever they want… this is a situation stemming from a personality dominance of a particular type of thinker. They can select from the population other kinds of thinkers and put them in charge- of the factor, of logistics, of marketing. It takes time to build a solid, healthy system- but it also takes time to kill of bad system… however, the first and most important effects can happen almost immediately… better quality goods offered to the people in the right locations, who have the freedom to spend a little extra to use them.

I suspect the inherent problem of communism isn’t it’s economics, but how it used it’s brain power… had they taken a earlier approach to thinking styles (they did eventually) they could of cleared up alot of issues and built a more stable and enjoyable system. Not to say I would of fought tooth and nail to keep them around.

5

Yep.

When you make people self-conscious, they take advantage (and become paranoid) of it.

6

What? Do you mean that they took 3 different tests of efficientcy? This is selfcontradicting.

7

How is using different measures contradictory? Don’t you ever do different things than you did before? Is it contradictory to eat steak today and spaghetti tomorrow? I don’t see why it’s contradictory.

8

You need to specify how and why making a messurement gives different results, even if it’s metaphorical.

9

I’ll spell it out for you if you insist, but you should know that most people find the example you quoted intuitively obvious.

If I’m your superior and I tell you I’m judging your performance based on number of nails produced, and you know, given say 8 hours, you can produce 300 small nails, 100 big nails or 80 super-big nail, what will you do? You’ll produce the small nails. Right?

If I’m your superior and I tell you I’m judging your performance based on the weight of nails produced, and you know that, given 8 hours, you can produce 1kg of small nails, 1.5kg of big nails, and 3kg of super-big nails, what will you do? You’ll produce super-big nails. Right?

There’s no metaphor involved. It’s all literal. It’s all just based on a basic understanding of human motivation. If I know that my performance is going to be based on X, I will maximize X. Otherwise…well, I’ll get a lower performance rating, risk losing my job or not getting a raise, etc. It doesn’t require some crazy theory about human motivation to understand this, just the basic idea that people want to keep their jobs, get good performance ratings, try to impress their superiors. If you’ve met people before, this should be intuitive to you.

10

Thanks F J for explenation.

What you are saying is that it’s human nature to be lazy and take the easy way out. But if you train people and allow them to do quality, they will do quality as best as they can, just look at germany, their saying is “ordnung muss sein” there must be order (and quality).

11

I don’t think it only extends to human nature. I think it can extend to a lot of areas in which one substitute goal is rewarded in an effort to maximize some other, more desirable goal.

12

I’m pretty sure profit is the reason nails today are of such poor quality. The issue is fixation on one goal at the expense of all others. For instance, if the only goal in the workplace is profit, people will hate their jobs and quit to become self-sufficient farmers, or the company will rape their clients for a short time before losing everything in court.

13

Desireable would still partially suggest it’s desireable to be lazy and take the shortest way.

14

You forget that it’s in old communist Russia, there’s not such thing as profit in relative terms, every company will have the same huge welfare support, no matter how bad or good their product was, therefore it usually was no sense in doing an effort in hard work.

15

I don’t see that as a profitable situation…
If I lost everything in court, I’m not sure to what extent I ‘profited’.

‘Profit’ is very broadly defined for me. If I trade 2 dollars for a hamburger that’s really good and takes away my hunger, I feel I’ve profited. Profit is usually seen too narrowly to be about money. Profit is about making decisions that you like the result of, more than the result of not making that decision.

Anyway, when it comes to higher quality nails, I’d imagine that’s more down to consumer demand rather than producer greed.

16

I don’t know what that means. Desirable for whom? What would suggest that? This was a very vague sentence.

17

People can get rich running companies into the ground. So the company can go bankrupt while the officers laugh all the way to the bank. And you won’t get any better nails than you do in a communist country.

But philosophically, my point is that the problem is narrow goals, not substitute goals. The conception of substitute goals is an interesting one, I’ve discussed it on ILP before myself. But I’m questioning whether the contrast between capitalism and communism has anything to do with it.

18

Well…that’s a bit hyperbolic. Last time I checked, Home Depot was stocked full of nails of a variety of sizes, all of which are usable. They may not be of as high a quality as you personally would want, but that’s different from saying ‘They’re practically unusuable’ as in the example.

The thing it has to do with is that the whole thing is about incentives to perform well. In capitalism (in fact, let’s not talk about capitalism but ‘free market’) the incentives are arranged differently. In a (non-corporate) free market, if I’m producing nails, I get money if people want to buy the nails. If my nails are too small, people won’t buy them. If they’re too big, people won’t buy them. So I have no incentive to produce nails that nobody wants. The only reason to make nails is if people want them and give me more money for them then I had to spend producing them. So I have to answer to the consumer, at the end of the day.

In communism, you don’t answer to the consumer, you answer to the guy who determines your quota. Now, what the real goal is is more complex than “I want a lot of nails.” If we’re trying to contribute to society, we want a lot of nails, but we also want a variety of nail sizes, and we want them usable, etc etc. But it’s easier to judge how many nails someone produces, then it is to judge if they’re producing ones that people want in appropriate varieties. So what the planners do is they use the easier measure, as a substitute for the real goal, because…well, it’s easier to measure. And so they end up incentivising the wrong things, because the nail manufacturers are answering now to them and their too-narrow goal (I agree with you about narrowness – that’s actually one of the proposed solutions to Goodhart’s Law - to use more measures if you can’t easily measure your real goal).

Communism is just an example. An example where the incentive was too-far removed from the real goal.

19

This isn’t about communism anyway. I’m not arguing that communism is bad, or that this is an inherent part of communism. I’m not arguing against communism. If it is a true story, I’m sure the person in charge of making those policies and setting those targets was just a not-so-bright fellow, and communists could indeed produce a metric for measuring nail production that would result in the kinds of nails that people want.

This isn’t about communism or capitalism.

It’s about goodhart’s law.

20

““substitute goal is rewarded in an effort to maximize some other, more desirable goal.””
Ok maybe i should ask is it desireable goal for the state, the factory or the worker?