trivialbenj: (Default)
[personal profile] trivialbenj
Ryan, Lori's friend who is now working at Bridgewater Capital in Stamford, made a very interesting comment when we were sitting around ordering tapas. He said that when he went out with people from work, this process of collective ordering never happened. "Food just got ordered," he said. And it's funny, because it was what I was thinking but just couldn't put my finger on. The biggest culture shock of my vacation was a movement from a top-down decision making process/environment to a communal or, as I kept thinking about it, a consensus based decision making process.

When I hang out with friends from work, the decision making tends to happen before we ever get out the door and it's done on an opt-in/opt-out basis. I'll call up a friend from work and say "I'm looking at going for Ethiopian food this weekend; you want to come with?". The implied question: would you like to get Ethiopian food [yes/no]. At the outset, there's kind of a broad expectation of what's going on. A bunch of us get to the restaurant and the first question is some variation on "family style or no?" If there's a bunch of people, mostly family style but excepting a couple of people is to be expected. If there are basically no dishes everyone wants to try, we go single plates. Consensus is maintained in the process, but mostly through giving numerous opportunities to opt-out with no social cost as the event goes on.

It's never stated explicitly, but the underlying mechanic is that someone is in charge. Someone suggests the cuisine, someone suggests the restaurant, someone suggests an ordering style, and other individuals either voice concerns or tacitly approve. (Approve through their silence, for those unfamiliar with the phrase. It's a good one.) It feels to me like when you get an assignment from your boss at work, you either jump to it or you express concerns at that point. I guess it's that I frequently assume when I'm surrounded by smart people that if one of them holds a strong opinion on a topic I'm indifferent to, I defer to them. They care more, they've probably given it more thought, etc. I'm not saying this is a 100% rule, but it feels logical to me.

Now, increase the social cost to opting out but maintain the importance of consensus. "We're all going to a bar, and you're coming whether you like it or not." It's now critically important for it to be the bar that you like. You want the drink that you want. You need to leave when you need to leave. Right? As this sense of "we" is imposed on you by your lack of opt-out option, you have to ratchet up the sense of buy-in to the group decision. But this can become complicated. The more people you have, the more difficult it is to pick one outcome that satisfies all of them.

This brings me to another problem. I'm pretty sure it's from The Paradox of Choice. There are two kinds of decision-makers on any given topic. There are "maximizers" and "satisficers". A maximizer wants the best car. He's going to look at all the cars on the market, optimize them over gas mileage, safety rating and price, pick the best location for pickup, and empirically determine the best color for picking up girls. A satisficer's mantra is "Is it good enough?" She'll try out a couple of different places, test drive a couple of different cars, and decide on one.

Introduce maximizers and satisficers to the last example. The more maximizers you're ordering with, the more likely you are that one person will say (implicitly, usually not out loud) "That's pretty good, but we could do better". Instead of the rest of the table saying "You know what, get whatever you want for yourself, Max [as our maximizer will be called for the rest of this example], but leave us out of it", the rest of the table does not want to be seen as excluding Max and Max does not want to appear anti-social. Plus, Max benefits from tasting many more dishes than he could order on his own, so he doesn't want to opt out. Imagine now that we have Max and Maxine, each of whom like different dishes.

Using a concept from computation here, think about what happens to the difficulty of an optimization as the number of choices grows. It's obvious for any group that increases the number of choices makes the decision harder, but increasing the number of choices for someone attempting to maximize makes the decision exponentially harder.

You might think I'm shooting myself in the foot pointing this out, but the more maximizers you add shrinks the universe of acceptable choices. Max doesn't want any fish dishes, Maxine is a non-drinker and won't have anything cooked in wine. Less choices, so better! What's the cost? Time and energy. XKCD, do my work for me! Picking out the best option takes time. A lot of time sometimes. In Boiler Room, they talk about "starting from yes". Ben Affleck gives the speech to his new charges about asking questions you know will get yes answers because they lead to more yes answers. If the implicit question is "Is this the best menu for me", you're starting from no. If the implicit question is "Is there enough stuff I like on this menu", you're starting from yes.

Discussing this problem with Katy earlier gave another interesting situation in collective decision making somewhat different from those above. Basically, it's the "I don't know, what do *you* want to do" scenario. When n gets big enough ("We can go anywhere for dinner"), it becomes one person throwing out ideas and other people shooting them down -or- nobody throwing out ideas and everyone just sitting there.

(I just noticed number of choices above is kind of ill defined. If the menu is n items long and we plan to order k of them allowing repeats, I would say that increasing n makes the problem much more difficult for maximizers and about the same for satisficers and that increasing k makes the problem more difficult for both. Though, I just did thought experiments for those, and I could actually see it going either way for satisficers on both cases. If n increases, a satisficer who didn't see anything good on the menu before could see a good new item. If k increases, a satisficer whose favorite dish didn't get ordered before can suggest it.)

So here's the tapas scenario. 12 of us go to a restaurant to get tapas. The party is no less than 1/3 maximizers, it's a birthday party (a traditionally bad opt-out situation), n is large (Tapas places have a big menu) and k is large (Tapas dishes are small). How do we solve this problem? Our actual solution: split into two groups. Although, God help me if getting people to move down so that people could sit with who they were ordering with wasn't a problem. It made the problem manageable, or at least moved the noise to a part of the table I wasn't sitting at. Ryan went with the flow for the first round of tapas and picked a dish he really wanted for the second round to eat himself, which was not a problem.

What's the general solution? There isn't one that I can see. Mostly, I just think the question is interesting. The one my family opted for (there are 6 of us: Mom, Dad, me, and my three younger brothers) was generally to stick to a smaller number of vetted options. We had places that we would go for family dinners out and a set menu of family dinners we'd have at home. It kept infighting low at home, but me becoming a foodie seems to be a rebellion from this tradition. Also, decision making was frequently deferred to someone in particular, because it's so-and-so's birthday or because we're in that neighborhood.

I don't know even if the basic idea is sound, but that's what I've been thinking about.

Also, the tapas plates aren't big enough for 12 people to get a taste of every dish. Even if everyone agreed, you couldn't get the dishes all the way around the table.

Date: 2007-12-13 05:30 am (UTC)
From: [identity profile]
It seems to me that if k/m (m the number of people) is >>1, as at a tapas place, then each person just orders k/m things he would like to eat (but everything gets shared anyway). This increases repetitions at the expense of marginal dishes, but makes everyone quite happy and is really simple.

Date: 2007-12-13 03:02 pm (UTC)
From: [identity profile]
You need k/m to be an integer. When you order 5 dishes for 3 people, how do you split up ordering?

Also, in general, people are reluctant to re-order dishes even if it's what they really like unless they also know it's a group favorite. Just an interesting social pressure.

Date: 2007-12-13 05:50 pm (UTC)
From: [identity profile]
5/3 is not >>1 :) At a reasonable tapas place k/m should be in the area of 3 to 5, depending on the person and the particular dishes, and so rounding to an integer is much easier.

Date: 2007-12-13 06:11 pm (UTC)
From: [identity profile]
Yeah, the real problem is when you're at a Chinese place and so k=m-1.

Date: 2007-12-13 05:52 am (UTC)
mackenzie: (Queen - Sunshine)
From: [personal profile] mackenzie
At the Palo Alto Poly dinner, many people eat family style. What this means is that you order what you want. When it arrives, you take as much of it as you want, then pass it around the table.

Many dishes don't make it all the way around the table, and that's okay. It just means that you see some shuffling as people move closer to those they believe will order dishes they enjoy toward the beginning of the meal.

Date: 2007-12-13 12:14 pm (UTC)
lorimt: (Default)
From: [personal profile] lorimt
I frequently go to dinner with the MIT Science Fiction Society. The group size varies from 8ish to 15ish, but dinner gets ordered very smoothly. This basically works because we usually go to the same few places, at which there are a handful of standard dishes and a handful of other known good choices.

Big groups (over ~8) get split up by table (resulting in 5+person groups), someone proposes a number of dishes to order (which allows people to opt out easily), and then the standards are put down on paper by someone, and a couple of alternates/duplicates are proposed.

I tried a variant of this at the tapas place, but we started out unable to agree on a number of dishes, then moved on to disagreeing about basic types of protein preferred when people started naming dishes (not to mention how many). This isn't a criticism of the people ordering, but of the system, for lack of a better term. Since there were so many dishes, and many unusual ones (and hearing was tricky), it was hard to guess who would like what, how popular something would be, etc. We pretty much had to split up, since tapas take a little negotiating to figure out what people want.

I think the general question of satisficing vs maximizing is a good one, but am not sure how well it works here. I'm mostly a maximizer, but my main issue was that the list of "animals I eat" only covered part of the menu, and unfortunately didn't match well with my initial seat-mates.

It's definitely an interesting question, though. I'm not sure what I think the answer should be for satisficers. For maximizers, it seems pretty straightforward that n going up is bad (more decisions to optimize over). As for k, there are more slots available to fit in someone's favorite dish, but more combinations to compare to one another, if you really want to find the best overall choice.

Date: 2007-12-13 03:00 pm (UTC)
From: [identity profile]
First of all, you have a really good tapas place. I remember being very impressed by that duck dish we had in the berry sauce and whatever pomegranate sauce we got from the other table. Plus, putting hummus out with bread instead of butter? That's just a genius idea.

For reasons not clear to me, though you hit some of them squarely on the head in your analysis, we kind of had a confluence of bad decision making elements all right together there. Breadth of menu was the biggest one I think I left off the list. If you're ordering a pizza, you're quibbling over whether you want varieties of pork or vegetables on your pizza (this covers all standard Domino's toppings except for extra cheese, cheddar cheese, ground beef, and pineapples). That menu was so broad that you could have essentially a seafood experience all the way to an exotic meat experience.

I also want to add that I'm using a really naive model for satisficing here. The spirit is true, but I don't know if there are actually people who say "good enough" that quickly and easily.

Date: 2007-12-13 05:12 pm (UTC)
lorimt: (Default)
From: [personal profile] lorimt
I think the whole thing would have been, if not a piece of cake, way easier if there were only 5-6 people or if the dishes were 1/2 or 2/3 the price, so people could order 3 dishes on their own and trade things away/plan to share a dish with one other person at most.

Date: 2007-12-14 10:23 pm (UTC)
From: [identity profile]
*I* think the main problem is our unclear definition of value.

But seriously, I think a significant contributor to this specific situation was that few of us (and none of our ordering group) had been to the restaurant before. Knowing what I know now, I would have gotten two of those duck dishes right off the bat, which would have probably taxed the limits of my clout and thus I would be done with the ordering.

However, I do think that the size of the dishes was unfortunate. One dish wasn't big enough to guarantee a taste for everyone, but two dishes wasn't reasonable unless someone actually liked it (which, again, goes back to the lack of a priori knowledge of the restaurant).

I know I was especially worried that a really tasty dish would be ordered, placed in front of a certain personage, who would wholly consume it before I ever got a chance to look at it, just less partake.

Despite my focus on the specific mechanics of this dinner, I am intrigued by the broader message of this post. The dynamics of communal ordering do seem to differ wildly from order to order and not necessarily along the axes you might expect. I'll have to keep an eye on such things next time I am insisting on even more chicken tikka masala.

Date: 2007-12-13 07:25 pm (UTC)
From: [identity profile]
I think most people are some combination: "I'll be fine as long as there is a vegetarian thing that is not spicy, but I really like olives, so if there's something with olives, I'll be happier. And I'm not fond of tomatoes, but I'll eat them if you make me."

Curse you, now I want tapas!

Date: 2007-12-13 07:23 pm (UTC)
From: [identity profile]
I'm in favor of the 'nearest neighbors' (or 'who's with me?') style of ordering at places where dishes feed only a few people and the group is bigger. Either you and the people next to you decide on a dish to share (minimizing passing dishes across a 20-person table), or people take turns shouting "I want the Foo Platter. Who's with me?" until they have as many takers as the dish supports. If there are more takers than the dish supports, the other set orders their own.

At places where it comes out to around one dish per person to fill people up, it works if everyone orders the dish they want most (or says "hey, I'm interested in these three dishes, which one do you like?" if they're a bad decision-maker like me). That way everyone is guaranteed to have at least one dish they like on the table (which is good for maximizers). At places like Buca, where size of dish > food for one person, 2(small)-4(large) people need to agree on a dish, and if they're all maximizers this can be tricky, admittedly. (also bad is when there are 3 people, 2 want one dish that the third can't eat, and dishes server 2-4 people! This happens to me a lot because I go out to 3-person dinners frequently and I can't stand spicy food.)


trivialbenj: (Default)

June 2010

131415 16171819

Style Credit

Expand Cut Tags

No cut tags
Page generated Oct. 19th, 2017 01:36 am
Powered by Dreamwidth Studios