Augmenting Group Intelligence

[TeresaJ]

We human beings find it extremely difficult to conceptualize concepts that are:

• complex
• dynamic
• partially opaque
• exponential

(borrowed from The Logic of Failure, which I read years ago)

We have developed computer systems capable of modeling complex, dynamic systems. Thinking of short term weather forecasts, which are still extremely simple compared to the enormity of society and culture (what I'm getting at) but nonetheless remarkably sophisticated compared to where they were a few years ago.

Nonetheless, in many respects, humans and the human subconscious are still smarter than most algorithms. Pandora and Amazon suggestions come to mind as being quite simplistic. "If you like x, try z" vs the human concept of "She likes x because of its a,b,c qualities, but only when she's in this mood or that context... so I intuit that, in this situation, she might like... e."

A computer can't make a "leap of logic."

Coming around to my question... How can we best use technology to augment human intelligence? And I don't mean on an individual basis (calculators, EverNote) but... something more like if WikiPedia could think - could store not just a narrative but several competing scenarios, a multitude of conditions.

This gets kind of into systems theory and informatics... There is some technology out there, like CogniScope, which is (was?) a proprietary decision-making software marketed mostly to large organizations.

The CogniScope methodology is a codified and tested means of collaboratively defining a complex situation and developing a social contract amongst the situation’s representative stakeholders regarding directives to address it (Christakis et. al, 1988; Warfield and Cardenas, 1994, Christakis and Dye, 2000). The
process imposes a structured discipline of “focused and open dialogue.” Stakeholders generate and clarify observations of the situation, collaboratively discern collective challenges, and construct patterns of interaction amongst participant’s observations. Principal outcomes are individual learning, integration of the diversity of viewpoints, the discernment of salient priorities for design, and the emergence and codification of a situation-specific consensual linguistic domain that enables communicative action.

Is anyone familiar with technologies like this? What do you think about the possibility of using technology to enhance human decision making?

[HilbertSpace]

This is a big chunk of what I do for a living.

Weather systems are (usually?) considered chaotic, rather than complex. Complex systems have more organization (to put it briefly - Stu Kauffmann wrote extensively on this, and I pretty much endorse his approach). Evolutionary systems (biological, technological, social, ...) are complex. Complexity theory grew out of chaos theory, and the two are related, but distinct. That's not really important, though, but the distinction might make it easier to find related papers and books.

I'm not sure how you're thinking about "leaps of logic," but for the sort of things you're talking about, computers can and do execute such leaps. There's still a lot of work to do (and thus I remain gainfully employed). The algorithms executed by companies like Amazon and Netflix are relatively simplistic, in a modeling-of-human-thought-and-behavior point of view. Some approaches are derived from the techniques developed in computational linguistics (where we're looking to discover concepts that are related to each other purely by looking at statistical relationships in texts), but applied to (say) book recommendations based on previously selected books. Commercial entities (and people interested specifically in the algorithmic side) aren't in general thinking about any deeper implications, and so try to optimize around fitting particular data sets, and are further constrained by having to convince a person with a checkbook that a possible incremental gain in fit is worth an additional investment. I'm fortunate in that I'm not (generally) constrained by things like that at this point.

In any case, there's a huge literature on the subject. I'm very familiar with the work of people like Christakis, but he's more about analyzing data than modeling systems. If you're interested in computational modeling, you might take a look at social physics as a topic area. I'd also recommend looking at Scott Page and Joshua Epstein. Those books are modestly technical, but they'd give a good starting point. I can try suggesting less technical books and articles, but I'm a lot less familiar with those. The general search terms, though, would be things like agent based models of social systems and complexity theory.

Edit: I think that Scott has (had?) a Coursera course on line. There are several courses on Coursera that relate to this sort of thing, and I think they'd be at an accessible level.

[TeresaJ]

Yes! Thank you! I will try to respond in more detail when I have had a chance to digest some of this. It might take a few years.

I wonder if anyone else here works in a similar field.

Also, I wonder if there are any programs that exist or are in the works that would make this sort of thing - I mean the application of it, not the concept of it - accessible and useful for non-specialists. Something to mess around with to get a feel of how it works.

[HilbertSpace]

Yes! Thank you! I will try to respond in more detail when I have had a chance to digest some of this. It might take a few years.

I wonder if anyone else here works in a similar field.

Also, I wonder if there are any programs that exist or are in the works that would make this sort of thing - I mean the application of it, not the concept of it - accessible and useful for non-specialists. Something to mess around with to get a feel of how it works.

I just read a cogniscope abstract. Totally different Christakis than the one I was thinking about. I think the stuff I mentioned might be more interesting and relevant, but for a good introduction to general systems thinking, I highly, highly recom!end Thinking in Systems by Meadows. It's written for lay people, and goes over system dynamics modeling. The field is very much about bringing systems models to decision makers.

[Starjots]

Coming around to my question... How can we best use technology to augment human intelligence? And I don't mean on an individual basis (calculators, EverNote) but... something more like if WikiPedia could think - could store not just a narrative but several competing scenarios, a multitude of conditions.

I highly, highly recom!end Thinking in Systems by Meadows. It's written for lay people, and goes over system dynamics modeling. The field is very much about bringing systems models to decision makers.

I added the bolds to help think about one aspect of the topic. Some questions this brings to mind:
Q1: Who is acting on the improved information?
Q2: Why are they doing so?
Q3: Who will be affected and will the affect be positive or negative?

A1: The most obvious application is for decision makers, which I assume means those who decide for others, CEOs or government leaders. But the people using these tools would more likely be the experts who make recommendations to the decision makers. I doubt high level leaders play around with social simulation tools.

A2: To improve the efficacy of their decisions.

A3: The same people affected by the decisions to begin with and it depends on the (1) quality of the tool (2) the intentions of the decision maker. Which is to say, such tools, only amplify the ability of the decision maker to turn their intentions into reality.

If the system is win/lose (e.g., the stock market), it amounts to an advantage to the players (decision makers) with the right tools. Those who don't have to tools tend to lose more and the tools themselves would be protected as proprietary information.

If the system is win/win (I'm having trouble thinking of such a human system other than those related to common survival problems such as disease, catastrophic global problems etc) then the tools would be apt to be shared and others encouraged to use them.

I gather from the OP that what is of interest in more in the win/win category since wikipedia was mentioned - everybody benefits.

--

Thanks for the book recommendation. This sort of thing is very interesting to me as well. Professionally I work in computer networking, which seems to be a related field in some ways. I'm interested in game theory and how that, combined with networking might yield better models for human systems. Is this part of the approach taken by the experts?

[HilbertSpace]

I added the bolds to help think about one aspect of the topic. Some questions this brings to mind:
Q1: Who is acting on the improved information?
Q2: Why are they doing so?
Q3: Who will be affected and will the affect be positive or negative?

A1: The most obvious application is for decision makers, which I assume means those who decide for others, CEOs or government leaders. But the people using these tools would more likely be the experts who make recommendations to the decision makers. I doubt high level leaders play around with social simulation tools.

A2: To improve the efficacy of their decisions.

A3: The same people affected by the decisions to begin with and it depends on the (1) quality of the tool (2) the intentions of the decision maker. Which is to say, such tools, only amplify the ability of the decision maker to turn their intentions into reality.

If the system is win/lose (e.g., the stock market), it amounts to an advantage to the players (decision makers) with the right tools. Those who don't have to tools tend to lose more and the tools themselves would be protected as proprietary information.

If the system is win/win (I'm having trouble thinking of such a human system other than those related to common survival problems such as disease, catastrophic global problems etc) then the tools would be apt to be shared and others encouraged to use them.

I gather from the OP that what is of interest in more in the win/win category since wikipedia was mentioned - everybody benefits.

--

Thanks for the book recommendation. This sort of thing is very interesting to me as well. Professionally I work in computer networking, which seems to be a related field in some ways. I'm interested in game theory and how that, combined with networking might yield better models for human systems. Is this part of the approach taken by the experts?

A1: You're right - most decisions don't go all the way up the chain. Even when they do, they're usually made by the people providing the people at the top of the chain with recommendations. However, even those people aren't so much the people actually *using* this kind of tool. System dynamics is a methodology for decomposing complex systems into interacting components. SD tools like VenSim pretty much provide a graphical framework for putting together ordinary differential equations. The SD methodology, though, also includes constructing a dialog with the actual stakeholders. So the modeler and the subject matter expert work together to design a model of the system, and the SME can then be informed by taking a systems perspective into account. The Sloan School of Business (MIT's B school) has a strong faculty around this sort of thing.

Game theory and network modeling and agent based models are a related area, and should be seen as an alternative, but complementary, approach to modeling and analysis. That's more what I do. The SD stuff is oriented around top-down analysis in a way that captures cascading consequences (if a+b=x and x+y=z, then changing a changes z). Game theory and network analysis, on the other hand, use a more bottom-up approach to try to understand the fundamental underlying physics of the smaller components that give rise to the large scale behaviors of the system.

I think that the understanding of networks (both from a formal starting point like graph theory) and a domain perspective (like information flows) is key to understanding these phenomena.