kitchen table math, the sequel: Crowds vs. herds

Saturday, June 4, 2011

Crowds vs. herds

In my book I draw a distinction between "cooperation" and "collaboration," defining the former as people working while interacting, and the latter as people working on joint projects, but not necessarily in one another's presence or with much productive interaction. In collaborations, after the work is divvied up, participants might spend the majority of their time working independently. 
I argue, furthermore, that this is what typifies most successful real-world collaborations. Except for those of us working on construction sites or film sets, we tend to get most of our work done at desks in private offices or cubicles; not at conference tables.

It turns out that there is a good reason for this. In an article in last weekend's Wall Street Journal, Jonah Lehrer reports that:
The good news is that the wisdom of crowds exists. When groups of people are asked a difficult question—say, to estimate the number of marbles in a jar, or the murder rate of New York City—their mistakes tend to cancel each other out. As a result, the average answer is often surprisingly accurate.

But here's the bad news: The wisdom of crowds turns out to be an incredibly fragile phenomenon. It doesn't take much for the smart group to become a dumb herd. Worse, a new study by Swiss scientists suggests that the interconnectedness of modern life might be making it even harder to benefit from our collective intelligence.

The experiment was straightforward. The researchers gathered 144 Swiss college students, sat them in isolated cubicles, and then asked them to answer various questions, such as the number of new immigrants living in Zurich. In many instances, the crowd proved correct. When asked about those immigrants, for instance, the median guess of the students was 10,000. The answer was 10,067.

The scientists then gave their subjects access to the guesses of the other members of the group. As a result, they were able to adjust their subsequent estimates based on the feedback of the crowd. The results were depressing. All of a sudden, the range of guesses dramatically narrowed; people were mindlessly imitating each other. Instead of canceling out their errors, they ended up magnifying their biases, which is why each round led to worse guesses. Although these subjects were far more confident that they were right—it's reassuring to know what other people think—this confidence was misplaced.

The scientists refer to this as the "social influence effect." In their paper, they argue that the effect has grown more pervasive in recent years. We live, after all, in an age of opinion polls and Facebook, cable news and Twitter. We are constantly being confronted with the beliefs of others, as the crowd tells itself what to think.

This research reveals the downside of our hyperconnected lives. So many essential institutions depend on the ability of citizens to think for themselves, to resist the latest trend or bubble. That's why it is important, as the Founding Fathers realized, to cultivate a raucous free press, full of divergent viewpoints.
The ideal, then, isn't group think, but independent thinking followed by a compilation of people's thoughts. 

Jonah Lehrer, however, neglects to mention one reason why the social influence effect has grown in recent years:  all the time that today's students are forced to work in groups in K12 classrooms, and, increasingly, in college classrooms as well.  In this case it's not the hyperconnectedness of our wired and wireless lives that's responsible, but the group think of the education world, with its systematic confusion of "cooperation" with "collaboration."

(Cross-posted at Out In Left Field).


Bonnie said...

Companies are increasingly moving to the crowd way of doing things. In the software industry, it is rare to find private offices. I visited Google Manhattan recently. It is all open work spaces, with people migrating from place to place with their laptops or tablets. They do have cubes, but the walls are very low so people can see each other. At the last company that I worked at, we were in cubes, with software developers, support people, project managers and QA all randomly distributed, supposedly to increase team spirit.

So, maybe K12 should focus more on preparing students for a world in which private offices and private work no longer exists.

Anonymous said...

For what it is worth, my company is (or was ...) aware of this effect.

Two examples:

Our procedure for building schedule estimates was for a handful of people to provide estimates for the different parts of the schedule. Only after we had the estimates, did the people involved talk them over with each other. The idea was to avoid "groupthink." It worked pretty well. The schedule estimates were fairly good.

When interviewing candidates, we try hard to avoid discussing the candidates until everyone has talked to them and formed an opinion. Only then do we try to reach agreement on hire/no-hire. Again, the idea is to avoid biasing an interviewer who hasn't talked to the candidate yet.

-Mark Roulo

Katharine Beals said...

"Google Manhattan recently. It is all open work spaces, with people migrating from place to place with their laptops or tablets. They do have cubes, but the walls are very low so people can see each other. "

Very interesting, Bonnie. Thanks for sharing. I would love to know more about how widespread this trend is--and also how employees feel about how it affects their productivity.

debbie stier said...

My personal experience was that Group Think was different than open offices, and it may be connected to a strong leader. When there was a strong leader, I noticed that people would take a stand, and experiment, and were more creative. When the fear set in (possibly by a weak leader, but also by economic conditions), no one would dare to think on their own, or take responsibility for any decisions, and everything began to seem watered down into the lowest common, unthreatening denominator.

I absolutely see this type of behavior in the education world too. And actually, when I read that article in the WSJ, it reminded me very much of the board of ed elections.

Linda Seebach said...

The newsrooms I worked in were all open (it must have been a nightmare when everybody was using typewriters) and it seemed to be positive because people who often worked on the same kinds of stories were fairly close together and a slightly louder than normal comment about the difficulty of reaching somebody might bring a response from someone who had his cell phone number (and you couldn't just ask him for it, because you didn't know who it was). Sounds distracting, but less so than sending a message to the entire staff.

My son's computer company has people working in widely scattered locations (many at home) but of course they're all online all the time anyway.

Bonnie said...

We are actually working today on a project with people at Google, but we are in constant communication - we have a webcam feed set up, and are chatting both on Google groups and Skype every few minutes. The people on the site with me are all sitting together at two tables.

Katharine Beals said...

Bonnie, is there a point at which you will divide things up and work individually, or is the entire project (or the majority of it) completed during this constant communication?

SteveH said...

Just because cube walls are lower (and email instantaneous) doesn't mean that people are working on the same task at the same time. There has to be division of labor at some point. If you can't clearly define the boundaries of individual work, you will have lots of problems. Also, the smaller the tasks and the more boundaries you have, the more problems you will have. There is a reason why "The Mythical Man-Month" has been a classic for over 35 years.

There are all sorts of ways to approach large complex problems from a system analysis approach; top down, bottom up, inside out, or outside in. You can have a rigid Analysis, Design, Code, and Test (unit and integration) process, or you can have a prototyping approach. All methods, however, require the work to be broken into individual tasks with clearly-defined boundaries. I have lots of WebEx meetings (they waste time), but that doesn't mean that there are no clearly defined individual boundaries. In fact there is a hierarchy of boundaries between the customer, our project, and the individual tasks.

In schools, where the goal is to work in groups without any leader or sage on the stage, the division of labor (if it is done at all) is decided by a group of equal members, unless one person decides to take charge. That doesn't work well if you have two or more strong personalities. This might be a learning process, but it's not what's done out in the real world.

Also, in schools two people can be working on the same thing at the same time. This is inefficient unless one person is training the other. Usually, kids will figure out how to split the work or one student will decide to do much more because they don't want the result to be screwed up.

Lots of noisy activity (active learning) in the classroom and lots of communicating in the workplace aren't necessarily good signs. Equality of workers in making decisions is also not necessarily a good sign.

Sombody should be in charge. That doesn't mean that one person is telling the rest what to do. You can have open "blue-sky" meetings, and open review meetings, but somebody has to make a final decision. However, for the majority of the time people should be working independently. This may not be true for managers, who tend to go from one meeting to another, but that is a different issue. Their job is to do this.

" estimate the number of marbles in a jar, or the murder rate of New York City—their mistakes tend to cancel each other out. As a result, the average answer is often surprisingly accurate."

Making decisions is different than working on a task. I would not like a company where decisions were made by voting. Schools encourage making group decisions (with no leader or sage on the stage) and they expect people to work together on the same task. The smart kids figure out that neither of these things work.

Bonnie said...

In our current sprint, we are all pretty much working in small groups at this point. Four to five people clustered around one computer...

What do you think of agile methodology (XP or Scrum), which has taken the software industry by storm? In a pure agile shop, you have a smallish team of co-located developers and testers, in an open space. There isn't a lot of design work. Instead, the project is divided into user stories that can be implemented in sprints of about 2 weeks. Everyone works pretty much together, but in particular, all programming is done in pairs. You always have two developers to a computer. And there is no task ownership - you can work on ongoing task A today, and tomorrow you might work on ongoing task B while Jane and Fred move to task A.

The traditional systems analysis model that Steve H described has pretty much died out at many software companies, and agile has taken over.

SteveH said...

Systems analysis is a general term. I was not describing one "traditional" form. I tend to use an outside-in prototype technique. Although having taught a course in systems analysis, I'm skeptical about almost every new form that comes down the software engineering pike, especially ones that expect process will overcome the need for good people. (I've seen a lot in the last 35 years.) The key advatange to Scrum is that it is a form of prototyping using a small group. The advantage is NOT because multiple people are working on the same task at the same time, unless you're stuck with weak coders.

Catherine Johnson said...

So, maybe K12 should focus more on preparing students for a world in which private offices and private work no longer exists.

I would rather schools not do this because in my experience schools constantly use 'the way it is in the real world' to justify anything they want to do at all, regardless of whether students are learning.

As usual, I'll add that I'm happy for parents and taxpayers to have the schools they want. If some parents want grade schools that require children to work in groups because major companies have employees work in groups, fine.

But I want to be able to place my child in a school that is focused on knowledge - and I want to pay taxes to support such schools.

And I don't want to be forced to put my child in a school where administrators assume they know how things are done at Google and thus are able adequately to simulate how-things-are-done-at-Google by putting everyone in pods or having their parents drive them around on weekends so they can work in groups.

This brings up an issue I see everywhere in K-12 philosophy, which is the notion that schools should be simulacra of adult life and adult expertise.

So much of the thinking about how to teach any subjects is predicated on the belief that one should find out what experts do and then have children do what experts do, too.

So, if experts do experiments, 6-year olds should do experiments.

If expert readers 'make predictions,' 6-year old readers should make predictions.

And so on.

My district's writing curriculum now has Kindergarten children do "research." They can't read or write, but they're doing "research."

Catherine Johnson said...

When there was a strong leader, I noticed that people would take a stand, and experiment, and were more creative.

That's interesting - !

These are subordinates to the strong leader, right?

Catherine Johnson said...

Haven't read the article yet, but this reminds me of research showing better decision making in groups where there is one dissenter.

I was reading some of the research the other day; it's a hoot.

These were studies where you show the subject two lines, say, (I think the stimuli were two lines), which are markedly different in length, and you ask him whether one line is longer, shorter, or the same length as the other. (I'm making this up - it was something like that.)

If every other subject says the two lines are the same, the subject says the two lines are the same, too, even though his private opinion is that one line is a lot longer than the other.

When researchers interview the subject, he or she says that the reason he agreed with the other people was that he thought he must be wrong.

When you change the experiment so that there is one other person who disagrees with the consensus, the subject now gives his real opinion.

Interestingly, when the other dissenter gives the wrong answer, it doesn't matter: the subject gives the right answer.

So if you have two lines of markedly different length and one person says that the long line is actually the short line while everyone else says the lines are the same, the subject will say correctly that the longer line is the longer line.

Catherine Johnson said...

ok, have read the article....

I think there's a distinction between 'wisdom of the crowd' and the 'social collaboration' of experts (or of people who are reasonably expert).

Speaking entirely from personal experience, I often find that two heads are better than one.

For my money, what people at Google are doing is quite different from what people asked to guess the number of immigrants are doing....

Again, speaking entirely from my own experience, when I'm working with a partner on something, I often know what I think -- and am trying to figure out whether what I think is correct, whether there's a piece of the puzzle I'm missing, what the next step is, etc.

I guess I'm talking about problem-solving, which for me is social.

I remember years ago reading that there's a difference between....scientists, I think it was....there's a group who work alone and a group who prefer to work in collaboration (or who need someone to brainstorm with).

At the time, that stuck me as plausible.

I'm in the need-someone-to-brainstorm-with group, which is one of the reasons I've written ktm all these years: I write ktm so I can read and think about the comments.

jtidwell said...

I worked at Google. I was based in the Cambridge (Mass.) office, but its architecture is comparable to the NYC office. I spent my share of time working as a visitor in the NYC office, too, and also in Mountain View.

What they are doing at Google is TOTALLY different from guess-the-number problems. :-)

But, first, the architecture. There are large rooms with an open layout, but these aren't generally huge. Most that I visited only held about 20-30 people, and the people working in them were respectful about a quiet environment. There are countless meeting rooms, both large and tiny, that one can retreat to when necessary. And truth be told, I never spent much time at my desk! I was frequently working in one of these meeting rooms, either in F2F meetings, or on videoconferences, or some hybrid thereof, or by myself when I needed serious quiet.

Plus, everyone is issued laptops instead of (or in addition to) one's desktop computer. Googlers are encouraged to take their work anywhere they please, as long as it's secure. Working from home is fine, too.

The nature of the work there is quite collaborative, as all good software product teams are. But once decisions are made, or questions formulated, everyone scurries off and does individual work. Very intense individual work. (In my case, it was UX design, with occasional programming or online research to figure something out.) And let me tell you, if you weren't extremely competent at self-directed work and individual achievement, you didn't last long in that environment. :-)

Following the crowd's opinion, just because you don't have the nerve to state your own differing opinion, isn't looked well upon either.

Catherine Johnson said...

What they are doing at Google is TOTALLY different from guess-the-number problems. :-)