> The scalar fallacy is the false but pervasive assumption that real-world things (hotels, sandwiches, people, mutual funds, chemo drugs, whatever) have some single-dimension ordering of "goodness".
> When you project a multi-dimensional space down to one dimension, you are involving a lot of context and preferences in the act of projecting.
Single dimension fallacy might be a good alternative name.
I'd add that often things don't have a single ordering of goodness at all. Like a tomato sauce: some people like it thick, others watery, others chunky; so it actually has at least 3 orderings.
Well, the general solution of economists is to assume that every person has their own ordering. This doesn't work for Spectrum Arguments by Larry Temkin, though. A typical argument goes like this: Suppose alternative A gives you an extremely high level of well-being for a month. B brings a little bit less well-being than A for two months. C a little bit less well-being for four months, and so forth. In pairwise comparisons you judge that B is better than A, C is better than B, and so on. Yet at some point, say, Z you will consider A better than Z if you compare A to Z, but you will continue to consider Z better than Y. Therefore, "better than" is not transitive.
People have written a lot of articles on how to solve these types of puzzles, which can be formulated many different evaluative domains. They always involve two or more criteria.
I've heard Temkin give talks attacking transitivity and I never buy it. Most of his examples involve infinity which is problematic. In the example you share, I think all it points out is that our ability to compare things (psychologically) is broken: we may think A < B < C < ... X < Z in pairwise comparisons, but something goes silly when we compare A and Z and think A > Z. I strongly think this is a bias akin to scope insensitivity. By the time we get to Z we're likely talking about HUNDREDS OF YEARS of experience, and we just are not psychologically fit to make such a comparison.
So I don't think Temkin is right.
PS - Derek Parfit and the "repugnant conclusion" ... not so repugnant if you think about it right ;)
That's a common reply. However, you can simply adopt these examples to fit whatever intuitions someone has, e.g. give Z a duration of thirty years and adjust the other intervals accordingly. Or give it a duration of 10 years and start with 30 minutes. Likewise, you can change the levels, replace well being with money or health or pleasure, and so on. Bear in mind that Temkin only needs one counter-example. I personally think we're psychologically fit to make these comparisons and that his arguments go through. It's not very surprising, if you take into account that in multicriteria decision making cases similar to these have been analysed and discussed since the 60s, e.g. by Peter C. Fishburn and by the French/Belgium tradition.
You can solve these without giving up transitivity. Technically, even just incompleteness solves the issue. My own favorite solution is lexicographic, but apparently not many reviewers like it. :(
Thanks for more comments. You're right you can re-scale things to be (A) 30 minutes VS 10 years (Z) ... at which point, why would anyone prefer A > Z ???
Part of the argument relies on this very vague definition of the eventual state of things. Temkin talks about an experience "just above 'barely worth living'" (Life-Z-1) ... and that is _so_ vague, and it means something _so_ different to different people. Given that people in concentration camps chose to continue living rather than kill themselves right away might make someone think "Life-Z-1" sucks, I would take 10 minutes of pleasure over that torture. Meanwhile I'm thinking what is just above 'barely worth living' (Life-Z-2) is just great -- the kind of basic enjoyment you have day to day without too much humor or excitement.
Philosophy is hard - I don't mean to poo-poo Temkin, I just never could grasp why transitivity was what he was attacking. As he says (something akin to) "there are several balls we're juggling in the air, and we must drop at least one to make it work" and he chooses to drop transitivity. And I don't see why he chose that 'ball'.
I can certainly see a case. Let's make it about money.
A is a single payment of $10,000.
B is two monthly payments of $5250. Or, in other words, a 5% improvement over A in total payments.
C is three monthly payments of $3675. Again, 5% better than B in sum.
I could completely see someone doing a pairwise comparison between A and B, and picking B for the 5% increase, because the time difference is not that important. Same again for B and C; it's only one more month for an extra 5%.
However, I could also see that same person comparing A and Z and deciding that instantaneous vs a period of two years is not worth the wait.
If a person is wealthy and doesn't need to spend that amount during the entire duration of Z, since they have other wealth to spend, they would likely choose Z for the higher return over the same time period.
So only a relative poverty would encourage the person to choose A.
Assuming they have unlimited life span. If not and their lifespan is uncertain, then the value of Z is reduced by the chance of not surviving long enough to profit from it.
If they can invest A and get a higher return than waiting for Z, then A is a more profitable choice.
Unfortunately time is one-dimensional, which creates a demand for ordering (I wouldn't be surprised if all demand for order actually came from that fact). I may not need ordering in 3D space, up until the moment I need to grab a few things - then I have to decide on an order, because grabbing is sequential.
And in all 3 cases you have putrid to fresh to masterfully made in terms of composition, quality of ingredients, age, etc...valuation can't be the composite of 'goodness', or we'd all become utility monsters chasing bitcoin.
A 3d vector of 2m by 8m by .1m can be projected on to a plane in a wide range of scalars.
When someone says that someone else is smart because they got a 100 on a test they are taking a multidimensional attribute and projecting it onto a single dimension.
I still think people understand that intelligence is multifaceted. Nobody past 3rd grade thinks that intelligence is entirely and directly measured by an exam.
However, acing a tough exam does show that you have high ability in some skills which can be included in "intelligence". It is perfectly reasonable to say someone is smart after they do something to prove their cognitive ability, even if they haven't proven every single possible kind of cognitive ability.
I think the miscommunication here is what people mean when they say "he is smart" or "this hotel is better than that one". I don't think people are commiting a scalar fallacy when they say these things. People are making judgements based on the many different aspects of a thing, and then using simplified speech to express their opinion.
If you are booking a vacation, you have to choose a hotel. When comparing hotels you look at all the different aspects (the bed, room size, the restaurant/bar, etc). You have to weigh all these options and then pick which hotel you would prefer to stay at. So when you make a choice and say 'hotel A is better than hotel B', you are not asserting that hotel goodness is measured in one dimension and that hotel A is objectively better. It's rather a benign statement that you would prefer to stay at hotel A.
It's not a fallacious reduction to a scalar quantity, but at some point you do have to make choices and weigh preferences.
> The scalar fallacy is the false but pervasive assumption that real-world things (hotels, sandwiches, people, mutual funds, chemo drugs, whatever) have some single-dimension ordering of "goodness".
> When you project a multi-dimensional space down to one dimension, you are involving a lot of context and preferences in the act of projecting.