Thus, education, especially science education, from Kindergarden through post-doc and beyond, should be organized around collaborations, teaching people and letting them practice the networking skills and collaborative learning and action. Individuals will make mistakes and get punished by the group (sometimes as harshly as excommunication). They will learn from that experience and become more collaborative next time. The biggest sin would be selfish non-sharing of information.
Hmmmm. On the one hand, yes: students can learn a great deal from one another, more than they can learn in isolation, and it's important for scientists (and lots of other people) to learn how to collaborate. But on the other hand, groups don't seem to work socially in the way that Bora imagines that they should.
I've observed many of the same things that Bora's commenters have: some students slacking off and letting others do the work, and some students getting impatient and doing everything themselves. Some of the students don't learn whatever their supposed to be mastering - and, yes, that is a problem. The class isn't going to go out into the world as a cohesive unit - they're going to disperse, and each individual is going to need to be able to apply the skills and knowledge as a member of a new group. (Would you want a member of a mapping team who couldn't identify rocks, or couldn't tell a bedding plane from a joint?) In order to work well in a group, an individual needs to be able to do his or her part.
So: if collaboration is good at some level, but students also need to individually master material, what's a teacher to do?
The simplest answer is to let homework be collaborative, but give individual exams. But I like to give students individual feedback when they're learning, before they're tested. That goes for the social aspects of group work, as well as the understanding of the material - I want to encourage the students to work well together while they're working on an assignment. So I've got a lot of different ways to try to judge mastery and encourage students to work together well while they're collaborating.
Grade based on reasoning rather than on answers. When I started teaching Structural Geology, I had a lot of field trip assignments with a lot of short-answer questions. I found that the weaker students would follow the stronger students, and both would write the same answers. So I started making students write papers for their labs, instead. When the students had to describe the outcrops in words, and then interpret their data, they were more likely to think the questions through for themselves (while also discussing the problems with the other students). Yes, they could have turned in identical papers - but it was a small class, and they knew that would be wrong. (I repeatedly lay out the rules for acknowledging help vs writing a paper together.)
Papers are hard to grade, but there are other ways to look at students' reasoning as opposed to just the answers. Concept sketches can be good. (For an explanation, see the materials for Steve Reynolds' presentation here.) Even just questions like "what's your evidence?" can push students to put the pieces together.
Make each student responsible for a critical piece of the project. This is the way real collaborations work. The jigsaw technique is great for small projects. For bigger projects, it's worth choosing a project that requires the effort of everyone in the class.
Have the project be something they want to be proud of. It's hard to figure out some kind of carrot beyond a mere grade, but if you can find one, it can be a great motivator. My writing class is in charge of putting together the alumni newsletter for the department, and if it isn't good enough to show to the world, I don't send it out... and everyone knows that the class blew it. I've only had one inadequate newsletter - knowing about that one class is enough to push all future classes to work hard.
Evaluate individuals orally in class. I've never done this, but Eric Baer of Highline Community College in Washington suggested this to me as a way to reduce the burden of grading labs. What if I circulated around the lab room while students were working, asking individual students to explain their reasoning as they did problems, and graded them on the spot? It's an intriguing idea, though I worry that I would have trouble getting to all twenty students, especially on the last problems.
Make critiques part of individual work. My structural geology class is currently working on a project that I fondly call "the cross-section from hell." It's a fairly straightforward problem, really: they've got a map of part of the Wyoming fold-and-thrust belt, and they have to draw a cross-section down to Precambrian basement, and it has to be reasonable geometrically. Each fault has to have matching beds in the hanging wall and footwall, and each bed has to be offset the same amount (unless there is a fault-propagation fold), and faults have to meet in reasonable ways. But it's really, really hard to make it work, and the solutions depend on the choices the students make at the beginning. (Which bedding orientations are most reliable? When should they average orientations, and when should they split them into two groups? What the heck is going on with that overturned fold, and why doesn't it follow John Suppe's rules for fault-propagation folds?)
After they've worked for a week, I have them exchange cross-sections with a classmate and look for things that don't work. They learn from one another, but they can't really cheat, because I grade each cross-section on whether it works, not on whether it matches my key. (I guess a student could go to the light table in the department office and trace a cross-section made by a student during a previous year, but that would take a lot of nerve to do under the eyes of all the faculty members!)
What about the rest of you? Do you have ideas that I haven't thought of?