LESSWRONG
LW

Comment Permalink

10y10

Optimizing assumes that you know you are moving into the right direction.

Eliezer recently wrote on FB in the LW Group:

I hypothesize that the best result will come from making up numbers, multiplying them, and then tossing them out the window and going with what seems like the intuitively best choice afterward.

As far as I understand CFAR also doesn't advocate doing things that feel very wrong on an intuitive level. To the extend that you believe that shutting up and calculate is useful, you haven't provided an argument for why you believe it's an optimization.

Gleb_Tsipursky

10y00

I'm confused by your presumption that I suggested doing things that feel very wrong on an intuitive level. Can you please highlight to me where I stated that? Thanks!

See in context

12 Personal story about benefits of Rationality Dojo and shutting up and multiplying

by Gleb_Tsipursky

26th Aug 2015

1 min read

12

My wife and I have been going to Ohio Rationality Dojo for a few months now, started by Raelifin, who has substantial expertise in probabilistic thinking and Bayesian reasoning, and I wanted to share about how the dojo helped us make a rational decision about house shopping. We were comparing two houses. We had an intuitive favorite house (170 on the image) but decided to compare it to our second favorite (450) by actually shutting up and multiplying, based on exercises we did as part of the dojo.

What we did was compare mathematically each part of the house by comparing the value of that part of the house multiplied by the use of that part of the house, and had separate values for the two of us (A for my wife, Agnes Vishnevkin, and G for me, Gleb Tsipursky, on the image). By comparing it mathematically, 450 came out way ahead. Hard to update our beliefs, but we did it, and are now orienting toward that one as our primary choice. Rationality for the win!

Here is the image of our back-of-the-napkin calculations.

Shut Up and Multiply

Personal Blog

12

Mentioned in

New Comment

26 comments, sorted by

top scoring

Click to highlight new comments since: Today at 2:24 AM

[-]Raelifin

10y50

Great job, you two! Don't forget to give your elephant and rider some time to "discuss" the findings internally before making the final judgment. I find that my elephant will slowly come around unless there's something important I've overlooked, which is a major risk when doing explicit calculations. For instance, I notice there's no representation of location, which tends to be a very important factor in deciding where to live.

[-]Gleb_Tsipursky

10y20

Thanks for the praise!

The location is pretty much the same for both places so we ruled it out. Appreciate your attention to this point, though.

Yeah, my elephant is slowly coming around as well. I myself was pretty surprised that the first choice lost out, it's a factor of many smaller diffuse points about the second-choice house combining to outweigh the couple of really big nice points about the first-choice house. Mathing it really helps deal with attention bias :-)

[-]tim

10y40

Setting aside emotion and simply doing the math is certainly worthy of praise in of itself. But I feel this anecdote would be better served after you have gone through with the purchase, lived in the house for a period of time and been able to say unequivocally: "I am really happy that we live in this house."

The pending uncertainty over the actual outcome casts a pretty big shadow over "yay we did rationality!"

[-]Gleb_Tsipursky

10y10

I prefer to avoid outcome bias, and focus on optimizing good decision-making processes :-)

[-]Luke_A_Somers

10y10

Yet, a good decision-making process will tend to yield better results. Have some confidence!

[-]Gleb_Tsipursky

10y00

Thanks!

[-]ChristianKl

10y00

Without looking at outcomes of a decision-making process it's hard to know whether it's optimized.

[-]Gleb_Tsipursky

10y00

Which is why I said optimizing, not optimized :-) Externalizing the decison-making process and then evaluating it using numbers is a way of optimizing it, not necessarily meaning it is the perfect optimized decision-making process. I decided to share my story right now because I thought the benefits to others in the LW community of me sharing it right now would higher than me having lived in the house for a while and then sharing it. Besides, by the time I live in the house, post-factum justification would be playing a potentially confounding role.

[-]ChristianKl

10y10

Optimizing assumes that you know you are moving into the right direction.

Eliezer recently wrote on FB in the LW Group:

I hypothesize that the best result will come from making up numbers, multiplying them, and then tossing them out the window and going with what seems like the intuitively best choice afterward.

[-]Gleb_Tsipursky

10y00

I'm confused by your presumption that I suggested doing things that feel very wrong on an intuitive level. Can you please highlight to me where I stated that? Thanks!

[-]ChristianKl

10y00

I don't think you suggested that thing it felt wrong but I think "shut up" suggests letting the data speak for itself and ignoring how it feels like.

[-]Gleb_Tsipursky

10y00

I was using the metaphor Eliezer used here, so I think it might be a semantics issue. The point of using math was to deal with attention bias, as I highlighted above. After that, it's important to evaluate feelings, for sure.

[-][anonymous]10y30

I'm curious how you avoided the typical human bias of not knowing what will make us happy in the future (IE, weighting your choices wrong).

I know you specifically have done a lot of research into what makes humans happy, so I'd be interested to hear how you applied that to this type of analysis.

[-]Gleb_Tsipursky

10y60

Easily enough, by modeling myself in a year from now and anticipating what my most likely preferences will be at the time. It's imperfect, of course, but it's quite a bit better than nothing.

Regarding making us as humans happy, I definitely factored that into my analysis. That's why I not only compared the value of each individual part of the house (kitchen, backyard, bedroom) but also the use of that part of the house. For example, the backyard is not really usable for half the year (not counting the view question, which was a separate component of the evaluation). The original first choice house had a gorgeous back yard, and both of our attentions was drawn to it. However, then we sat down and mathed it, and realized we would only enjoy it half the year, decreasing its total value by half. This is an example of how future-oriented thinking helped avoid a problematic decision that would undermine my total happiness.

[-]NancyLebovitz

10y10

Did you include the costs of adding features you like to each of the houses?

[-]Gleb_Tsipursky

10y20

Yes, we did, and both houses were about the same on this scale, so we did not include it in our calculations. Thanks for highlighting this point, though, as it would have been important if we missed it :-)

[-]Gunnar_Zarncke

10y10

Did you include subjective factors like neighborhood, location too?

[-]Gleb_Tsipursky

10y20

Yes, we did, and that's a point that Raelifin above mentioned as well. The houses are located on the same street, a couple of blocks apart. So this subjective factor was pretty similar, and thus we did not include it in our calculations. Thanks for highlighting this point, though, as it would have been important if we missed it :-)

[-]Luke_A_Somers

10y00

I did something similar very early in my recent house-buying process, trying to knock the list of candidates to look at down to something reasonable, because everything was a compromise.

... and then it was rendered moot because we got the house that had essentially all the attributes we wanted at once, at a price we could afford.

[-]Gleb_Tsipursky

10y00

Congrats!

[-]Jiro

10y00

What does multiplying the value by the use mean? That sounds like "I will use this X times, and each use is worth Y, so the result is X*Y." If that is what you are saying, that doesn't seem to assess the overall value very well because it is possible (and likely) that not all uses are worth the same; there is no fixed Y.

Of course, you could always figure out the total value first, then divide by X to get a per-use value. This value would work mathematically, but doing that doesn't get you anything, since you're really interested in the total value anyway.

[-]Gleb_Tsipursky

10y00

We had a 3-point system of value for each part of the house "1=low, 2=moderate, 3=high" and the same for use, with use averaged year-round.

So for the backyard example: the backyard for house 170 was high value, 3, and high use for half of the year, so 1.5. Its total value was 4.5. The backyard for house 450 was moderate value, 2, and high use for half of the year, so 1.5. Its total value was 3.

[-][anonymous]10y30

Save your calculations - you can go back later and judge how well you modelled your future selves.

My wife and I did something very similar 7 years ago when deciding between 3 flats, and recently came across our backs-of-multiple-envelopes calculations. We only considered the one we actually chose and estimated our current weightings for the different spaces. In our initial judgments we had vastly overrated our future use/value of the outdoor space, substantially underrated the kitchen and lounge rooms, and entirely ignored hall and loft storage space which has turned out to be decidedly non-negligible.

We are very happy in the flat so this is not an exercise in regret, hindsight or possible alternative universes - but made us notice that if we are flat hunting again, outdoor space is likely to cost more than it's worth to us.

[-]Gleb_Tsipursky

10y20

Yup, I'm saving them indeed, and this discussion post will be useful as we revisit our decision. I'll put a note in my calendar with a link to this post to go back and look at the post again in a year after we move.

Regarding outdoor space, I'm glad that my wife and I figured out in advance that we'll only use it half the year! Thanks for that feedback :-)

[-]Jiro

10y00

I was thinking of something that is only used occasionally, but for which you value the option to use it even if you aren't going to use it much. Because going from no-X to X adds the option, the first use has a high value, but the second use does not have as high a value.

[-]Gleb_Tsipursky

10y00

This can be calculated by estimating the probability of use times enjoyment, for example a formal dining room and table. Such a feature was not so important to us, so we left it out of our calculations. It's only salient things that matter :-)

Moderation Log