Re: Your Brain, How Does It Work?

1

We discussed some "how to (not) win an argument" research, but apparently it was different from the research you're describing.


Posted by: teofilo | Link to this comment | 06-28-14 10:00 PM
horizontal rule
2

I was going to say I think you're* all wrong again, but as I thought through the ways you could all be wrong, I came to the conclusion that maybe you aren't all wrong, after all.

*y'all


Posted by: fake accent | Link to this comment | 06-29-14 12:03 AM
horizontal rule
3

Hmm. Facebook have been sloppy and grasping in all sorts of ways before, but I feel like this one is genuine early adopter clumsiness instead of smug corporate indifference. That FB could manipulate emotional states with this sort of directness is not obvious, or not obviously different from say a newspaper.

Now it has been established with accidental transparency, ethics boards do seem the obvious needed step.


Posted by: conflated | Link to this comment | 06-29-14 4:23 AM
horizontal rule
4

The full first paper is here.. I might or might not buy it, not sure, but anybody looking for a vector of attack should note that it was a complex language task run on a relatively small number of subjects on Mechanical Turk.


Posted by: Beefo Meaty | Link to this comment | 06-29-14 5:17 AM
horizontal rule
5

The FB thing is probably unethical by the standard of university research, but it's hardly more unethical than what Facebook does constantly; showing you a subset of your friend's posts to some mysterious, possibly nefarious end is what they do.


Posted by: Beefo Meaty | Link to this comment | 06-29-14 5:19 AM
horizontal rule
6

I got tagged in my friends' posts because reunion photos.


Posted by: Moby Hick | Link to this comment | 06-29-14 5:37 AM
horizontal rule
7

I was wondering how IRB protocol works for smug corporations. What kind of process do drug companies have?


Posted by: heebie-geebie | Link to this comment | 06-29-14 6:08 AM
horizontal rule
8

A good article on the Facebook thing. Maybe not even as unethical as I conceded.


Posted by: Beefo Meaty | Link to this comment | 06-29-14 6:17 AM
horizontal rule
9

I'm mostly just deeply annoyed at FB's general withholding and manipulation of my feed. Just show me the people I've tagged as "show all", in chronological order.


Posted by: heebie-geebie | Link to this comment | 06-29-14 6:23 AM
horizontal rule
10

Right now, I keep re-seeing a post which cracks me up, which is a colleague of mine recounting the following incident: She meets an emeritus prof at Big Research U and they get to chatting. When he hears she teaches at Heebie U, he is floored and says something like "Our best grad students are INEVITABLY from there! I can tell without asking! You all must teach undergrads as though they're grad students! Rave rave rave!"

Obviously this is getting passed around like wildfire, but it really, really, cannot possibly be true. I believe my colleague had this interaction, but I wonder if the Emeritus guy had his colleges confused. Or if we only send 1-2 students per decade to his program. Or something. Or maybe I just never appreciated the non-math departments fully, because this is not the Heebie U I teach at.


Posted by: heebie-geebie | Link to this comment | 06-29-14 6:34 AM
horizontal rule
11

9: Like!


Posted by: Di Kotimy | Link to this comment | 06-29-14 7:30 AM
horizontal rule
12

9: I wonder what the game is that makes forcing posts into "Top Posts" order profitable. I have a theory that it's just Mark Zuckerberg falling in love with some algorithm he cooked up and trying to show it off at all costs. But really, is FB that stupid? Maybe.


Posted by: togolosh | Link to this comment | 06-29-14 7:56 AM
horizontal rule
13

But this is even with the "most recent" in effect. The "most recent" thing probably has some effect, but not the effect of actually strictly showing you things in the order of how recent they are.


Posted by: heebie-geebie | Link to this comment | 06-29-14 7:58 AM
horizontal rule
14

I noticed some problems along those lines. I rally don't care about the most recent comments on posts I didn't care about the first time I read them.


Posted by: togolosh | Link to this comment | 06-29-14 8:01 AM
horizontal rule
15

Indeed. And I don't need to see pictures of children of people I don't know just because one of my friends liked or commented on such pictures.


Posted by: Di Kotimy | Link to this comment | 06-29-14 8:33 AM
horizontal rule
16

12, 13: showing you what you want to see as efficiently as possible is definitely not what they're in it for. And whatever the strategy is, it has been subjected to tons of A/B testing (that is, psychogical experimentation) to see if it accomplished their goal, which is presumably whatever maximizes ad revenue.


Posted by: Beefo Meaty | Link to this comment | 06-29-14 8:36 AM
horizontal rule
17

12 -- I think it's ideological, yes. I mean, sure, they understand that some sort of hierarchical sort is necessary, or you end up with stuff like 14 and 15. But the machine isn't smart enough to figure out what we care about today.


Posted by: CharleyCarp | Link to this comment | 06-29-14 8:39 AM
horizontal rule
18

Help! Help! I'm being manipulated by a commercial enterprise!


Posted by: Turgid Jacobian | Link to this comment | 06-29-14 8:39 AM
horizontal rule
19

I've been running my own experiment with the trending box. How many times do I have to delete NBA stories giving "I don't care' as the reason before they stop showing me NBA stories? So far, it seems to be Halford's infinite overtime.


Posted by: CharleyCarp | Link to this comment | 06-29-14 8:42 AM
horizontal rule
20

You care enough to click on them. They've got your attention!


Posted by: Beefo Meaty | Link to this comment | 06-29-14 8:46 AM
horizontal rule
21

I could see how if you have multiple hundreds of active friends, some kind of algorithmic filter would be the only way to make the feed manageable. I started Facebook as a late adopter and have not really actively sought out friends, so my list is relatively small but even then it can take more time to catch up than I really have.

I'd still prefer to get all posts from everyone I asked for all posts from. It's annoying when something shows up a day or two later when someone I know comments on it and it's clear that lots of other friends have already seen it because they've already had a discussion, or there are a bunch of likes, or whatever.


Posted by: fake accent | Link to this comment | 06-29-14 11:06 AM
horizontal rule
22

I have a stock response to 9, which is along the lines of 21. At some point if you're just getting the "raw" stream - all of the posts from all of the people you're following, in chronological order - you'll fall behind. Maybe you're following too many people, maybe someone started posting more, maybe you were busy and did something other than check the stream for a while. What do you want to happen in that situation?

1. Nothing. Just see the same stream, and accept that you missed some things a while ago. If this is the steady state of affairs, accept that you're always missing things more-or-less randomly, based on when they were posted and when you happen to read.

2. See a "curated" stream of some kind, with some criteria applied to decide what's more important. There are lots of possible signals - number of comments, how closely linked the person is to you, how often you comment on their posts, etc. (Also possible: How much it makes you want to buy whatever is being shilled today). This way what you miss is the "less important" stuff.

I claim that 2 is better than 1. What annoys *me* is missing things that are filtered out when I do have time to keep up with it all, but that's less and less often the case.


Posted by: Nathan Williams | Link to this comment | 06-29-14 11:55 AM
horizontal rule
23

Sifu & others who apparently have more first hand knowledge of this study. Read the article, skimmed the study: I don't understand exactly what is supposed to be the distinction between explanation and reason-giving that is being invoked by the authors, nor how that distinction would be generalized to anything beyond a pretty narrow set of debates . There's a rough distinction I can imagine between explaining exactly how regulating industries of various sorts might have some effect on the speed of climate change (on the one hand) and why climate change is bad/has disastrous effects (reasons on the other hand). But even with some of the other debates with which the authors are concerned I begin to lose a sense of what this distinction is supposed to amount to-- e.g. merit-based salary increases for teachers could not be made to work appropriately (explanation) and that's a sufficient reason not to do that. (?) Then there are a whole host of hot debates, including political/policy debates, for which I don't have, really, any sense of how or why this distinction is supposed to help in discussing the issue. E.g. marriage equality. How does it work? well, let people enter into a certain form of legally recognized contract with one another irrespective of their sex/gender identity. Done. But the bigots don't disagree about *that*... ?


Posted by: backwardsinheels | Link to this comment | 06-29-14 12:12 PM
horizontal rule
24

On part 1 of the OP, how to win an argument: I love, love, love this sort of research. As noted in the previous thread on the topic, the linked article doesn't actually explain how to win an argument -- just points to making an effort to force people to explain themselves, which is of course the real problem. There's a hell of lot of resentment out there at being asked to explain / back up one's reasoning.

Vaguely related, a potentially nasty argument I had with a friend about a week ago. (We were watching the US-Portugal World Cup game, but that was not the problem.) Somewhere along the line, talk turned to US foreign policy, and my friend asserted that Shiites are Persian, while Sunnis are Arab, and that's why they hate one another.

Oh, no. After frowning and explaining that it was more complicated than that, that actually there are Arab Shiites as well, so the conflict occurs along two axes (religious and ethnic), my friend ... got upset with me, declaring that I was stupid if I didn't know that Shiites find the suggestion that they might be Arab deeply insulting, because they are Persian, idiot, and it's just this sort of American ignorance that fucks up our foreign policy in the first place.

Sigh. So I said again that I understood what he was saying about Persian/Arab sensitivity, but really, there are Arab Shiites. Really.

The point: he declared, upset, that Goddammit, I always do this to him, take a contrary position and insist in a superior fashion that he doesn't know what he's talking about. BUT I FOUND A WAY OUT: I just nodded, said that his certitude was making me doubt myself, that I was positive I'd read much writing contrary to his statements, and that we'd have to look it up later, because I did not want to fight at this time.

Totally worked. The following afternoon I had a couple of emails from him, the last of which was "I stand corrected." This has never happened before: he's always dug in. Guess I wasn't handling it properly.


Posted by: parsimon | Link to this comment | 06-29-14 12:25 PM
horizontal rule
25

The state of the art of sentiment analysis is just not that good right now. We have very little ability to recognize sarcasm, irony, or even simple negation in an automated way. So whatever size effect they claim to produce in the paper, I'd discount that quite a bit in your mental model of how well this type of manipulation works.


Posted by: dz | Link to this comment | 06-29-14 12:26 PM
horizontal rule
26

This, on the other hand, seems like a real thing (and a good thing) (and I'm not particularly a fan of Facebook).

http://ucsdnews.ucsd.edu/pressrelease/facebook_fuels_the_friend_vote


Posted by: dz | Link to this comment | 06-29-14 12:29 PM
horizontal rule
27

22: you forgot

3: give me the fucking tools to filter my feed however I want. (Or, at the very least, don't *actively fuck with my ability* to do so.) They do this to a limited extent with lists, but the way in which they hide them makes it obvious they don't actually want you to use them.


Posted by: Josh | Link to this comment | 06-29-14 12:39 PM
horizontal rule
28

Facebook has apparently removed the "show all posts from this person" option, so I guess we can't accuse them of ignoring it anymore. I haven't dug around Facebook's settings for a while. It really has become more like twitter, which has become more like Facebook. I thought people were exaggerating the extent of the convergence.


Posted by: fake accent | Link to this comment | 06-29-14 12:49 PM
horizontal rule
29

28: if it had become more like twitter, "see all posts from this person" and "give me everything in reverse chronological order" would be the defaults.


Posted by: Josh | Link to this comment | 06-29-14 12:52 PM
horizontal rule
30

The asymmetric Facebook "follow" is I guess a thing I was aware of, but hadn't really noticed. Twitter is starting to filter your notifications, but probably haven't fucked with you feed. Yet.


Posted by: fake accent | Link to this comment | 06-29-14 12:55 PM
horizontal rule
31

your


Posted by: fake accent | Link to this comment | 06-29-14 12:56 PM
horizontal rule
32

"Thick people are very good at winning arguments because they're too thick to realise that they've lost."


Posted by: Ginger Yellow | Link to this comment | 06-29-14 1:37 PM
horizontal rule
33

32: Interesting point.


Posted by: parsimon | Link to this comment | 06-29-14 1:53 PM
horizontal rule
34

OP.2 is relevant to my interests. I've not read that paper yet.

26: Unfortunately, the way they designed that experiment, it's hard to justify their conclusion. What they show is that people who have more treated friends are more likely to (say) they voted. (Actually they do a very scary record-linkage exercise to tie Facebook accounts to public voting records and estimate actual differences in voting rates.) But the problem is that when 98% of accounts are in the treatment group, having a lower-than-average number of treated friends pretty much means having a lower-than-average number of friends, period. And those people are odd, in ways which plausibly make them less likely to vote in the first place. The experiment they should have done was to pick a random person, and then pick a random friend, and then randomly decide whether or not to treat the friend.

I am told that at least one referee made this point to the authors, at a different journal, without effect. (What were they going to do, wait 2 years to run the experiment again the right way?)


Posted by: Cosma Shalizi | Link to this comment | 06-29-14 4:28 PM
horizontal rule
35

Please tell me there's a published study where the methodology is described as "trick or treat".


Posted by: fake accent | Link to this comment | 06-29-14 4:35 PM
horizontal rule
36

I think Facebook is conducting an ongoing experiment in whether people will repost anything that is vaguely meme-shaped. Right now on my feed someone has shared something that reads "Can you name a city without the letter 'a' in it? I bet you can't." Wtf drives people to share that? I guess whatever is driving me to post this comment. Shit.


Posted by: essear | Link to this comment | 06-30-14 12:00 PM
horizontal rule
37

36: modify it to "a city without the number '3' in it" and see if that spreads.


Posted by: Beefo Meaty | Link to this comment | 06-30-14 12:06 PM
horizontal rule
38

36: seriously.


Posted by: Annelid Gustator | Link to this comment | 06-30-14 12:11 PM
horizontal rule