In my experience/interpretation, people have a strong desire to predict/simulate because they then feel better in control to design and plan activities in order to to influence what is happeing. It's not prediction for it's own sake or pure contemplation (in that case, it would be enough to just wait till it happens), it's prediction in order to have a "sound" basis for designing appropriate interventions, or decisions (or to justify an already existing opinion).

Your points focus on the issue that it's not possible, and thus by false security even dangerous. While I agree, this argumentation leads to endless discussions (but it is possible - no - but a least a little bit - no because ...)
Another, optimistic, and possibly more pragmatic point of view is that it's not necessary, that you can save a lot of effort by not even trying. Instead, learn how to apply tiny, probing, diagnostic actions (and then sense how the complex system "resonates"), and iterate from that point onward.

It's not necessary to simulate or model the system in order to influence it.

Completely unrelated: I did not get the point with the postmodernists and your topic one. An insider argument?

Hi Christian
My point on PM is that some people faced with non-causal systems then lapse into post modernism thinking that because you reject empiricism you have to reject science

Dave --

You do not know what a relief it is to get some clarity in the conversation around complex systems. Whew. Thanks for your spot-on comments.

It is important to add that a very important compelex system for most of us, markets, does benefit from prediction.

The intellectual hurdle is that complex system prediction does not need to be causal or binary. For example, the contract price in a prediction market is simply the probability of a particular outcome. Most common is for it to range from 0-100.

Prediction markets are far more accurate that polling, top experts or even panels of experts.

Prediction markets need to be in the quiver of every complexity expert and KM leader to achieve mastery of crowd wisdom.

See: http://www.pmcluster.com/

Cheers,

-j

I'd classify prediction markets are a way to aggregate common opinions into a simple figure, motivating people by giving them a reward in exchange for their expertise. Prediction markets are "in", and it does not hurt to know about them, also first-hand, by participating (which I did in a real pilot for a buiness).
Problems include: The question must be framed properly; watch the rewards, why do people participate; it needs a non-trivial IT system; the market must be liquid enough (hence, sometimes there are agents active, set up by the sponsor, just to generate traffic); and my mantra: what's the bang for the buck: in a business context, coverting time of people participating (and there must be many since it has to be a liquid market) vs. value generated by the result: I have my doubts about the break even.

An a more abstract level, I'd say that a prediction market is a complex system in itself trying to emulate another complex system. There are non-negligible costs associated with this approach, and, since it's two different complex systems: it might work sometimes, but that does not mean .... (see David's 7 points).

Interesting - but systems do not exists and they cannot be defined. A system is simply one perspective on the world. As multiple perspectives always exist the trick is to try and come to a "common" definition. Howver this definition of a system will always be inperfect. Alternatively two systems always exist - the problem context system and the problem solution systems. So when we talk about a system which system are we talking about. Thirdly all system exhibite emergent properties and all systems can be defined in terms of thousands of attributes. Which attributes matter depend upon the onbe perspective. Howver all attributes of a systems whether defined or not defined can kill the system.

However all systems have multiple causes and rules sets hence complexity. I psoting above is correct but misleading in that it is true that western science have been based on linear cause and effect. Likewise I agree with the comment about six stigmata, howver the more enlighten companies incorporate compxity and multi cause into their six sigma programmes some 10/12 years ago incluing working not only on process variation but process variability and work content

Much of the original work on complexity was carried out some 15 years ago at IERC (Cranfield) sponsored by Honda

thoughts

Geoff Elliott

The words 'enlightened', 'six sigma' and 'complexity' do not often go together, at least not if by complexity one means a complex adaptive system. I know of work on simulation modelling which uses complexity principles in the context of manufacturing and I can see that this might be done. Anything involving human interaction to my mind is truly complex, limited in its ability to be simulated and not susceptable to the sort of measurement system required by the dictacts of six sigma. However I would be interested to hear of some examples - it might indicate some hope in the midst of much darkness.

Incidentally, there seems to be a trend of Executives implementing six sigma, achieving the normal short term benefits of any change and then moving on.....

- David: in my company, one attempt of motivating people to join the six sigma bandwagon is a very open statement "it's good for your career" ...
- Geoff: it does not matter if systems do exist ontologically or are just subsets/private perspectives to the (one) world: my point above was that if they are very different (like prediction markets vs. politics or box office slaes, or a business process map vs. the organization), you can not use one to RELIBALY predict the other. Thus the best model of a complex system is the complex system itself.

Hi Christian --

You made good comments particularly on liquidity and how the question or 'contract' is framed.

Prediction markets have been around for at least 100 years, first debuting in the political markets in NYC at the turn of the 20th century.

What's new and dramatic is Internet and Web mediation of markets. Prediction markets have hit their 'technology trigger' and are moving up the Gartner Curve. The scope or timeframe for the Gartner cycle is usually 10-years. Prediction markets will cycle through the curve much faster. They are quickly becoming routine and customary in the enterprise, for example.

Concerning the specifics of your comment, a couple comments/correction.

1.) You wrote, ..."prediction markets are a way to aggregate common opinions..." Not so. Perfectly orthogonal, because, in fact, and very specifically, prediction markets are a way to aggregate UNCOMMON opinions. If they were common opinions, there would be no market! (I think this is what you meant…)

2.) "...giving them a reward in exchange for their expertise..." is perfectly correct. In addition, there is solid research that non-monetary markets can do just as well and often better than cash markets, particularly in the enterprise.

Electronically-mediated markets are important, new. I recommend having a look at “Infotopia: How Many Minds Produce Knowledge, by Cass R. Sunstein, just released. See: http://kmblogs.com/public/item/140707

-j

John,
to your points
1. I agree with the common and uncommon in a way: yes, it's not just averaging what is in the middle of the road. My focus however was on the topic of aggregation: a mechanism to get rid of all the details of real people's expertise, beliefs, etc., in order to end up with a single binary answer (will the US invade Iran before December 31 2007 ?), or a simple, one-dimensional numeric answer (Movie X, what will be the box office results in the US for the first weekend).
2. Agree that there are more dimensions to rewards than money. The problem I have is that adding any reward is a distortion. In principle, you want a diagnosis (to tell you what is), not an intervention (to change what is). Thus the less influence the better, which runs agains the idea of rewards at all.

One the note of numbercruching/agregation: A while ago I read Mandelbrod's "The (mis)Behavior of Markets - A Fractal View of Risk, Ruin and Reward". I don't agree with everything he claims, but his point that the variability in markets is not gaussian but fractal seemed convincing. Which means that there are no intrinsic boundaries to fluctuations, another reason why the prediction may be misleading, and why averaging/agregation may not give you , reliably, what you want.

I think we'll end up with a statement of qualification: It may be good enough - within boundaries.

Christian

Microsoft: Use prediction markets to tap hidden knowledge. [Business 2.0]

A New Office Pool

If so-called prediction markets--betting pools in which shares are traded to gauge the odds of upcoming events--can call presidential elections and Oscar races with accuracy, why not use them inside companies to identify their next hit products? The answer: It's a lot harder to tap into collective brainpower about a product's market potential, and other key business questions, than it is to foretell the winner for Best Picture.

Microsoft, though, is trying to crack the code--and has been developing prediction markets as a serious alternative to blunt forecasting tools. Todd Proebsting, director of Microsoft's Center for Software Excellence, began running a series of prediction markets in 2004 to better gauge how many bugs a new software application might contain and to make more accurate calls about product ship dates. In his first effort, Proebsting chose 25 programmers and quality-control testers from a team of more than 50 working on a new Windows testing application. Via an internal website, workers could buy shares for the month they believed the product would ship. Shares were valued at $1 apiece, and the engineers drew on accounts stocked with $50 each to fund their bids. Within minutes of the site's launch, shares for a February ship date shot up in value while those for November, the scheduled release date, dropped to almost zero. That came as a shock to the project director, who had heard nothing but optimism in meetings and e-mail updates.

Exposing such communication gaps gives prediction markets even more potential value to businesses, Proebsting says. "Face-to-face communication breaks down, and opinions are filtered," he says. "Prediction markets show where the gap is and allow you to short-circuit it." In this case, the project manager did more troubleshooting ahead of time and avoided a delivery crisis. On the strength of those results, Proebsting ran another two dozen markets on a variety of software projects. Each, he says, accurately predicted the completion date. What's next, then, for this nascent technology in Redmond? Daniel Ling, chief of Microsoft Research, will say only that it's something the company "continues to explore." -- P.K.

--------------

Hmmn, '...continues to explore...' is managerese for, '...we want to kill it, but in a way that won't attract any attention...'

Prediction markets (and complexity science, btw) usurp command and control. So-called 'management' doesn't like 'em, but leaders do. Ironic that today's management craves predictability, yet shuns learning, crowd wisdom, complexity, etc.

It is just the normal pains of living on the bleeding edge.

Honestly, there are surprisingly few critics of prediction markets.

-j

Oh dear .. if your going down this route ..read Guy Claxton first .. "The Wayward Mind" and the antecedents

Very cryptic Mark!
I've read it but not sure if you are targeting John, Christian or myself?
Can you summarise the point for the benefit of us lesser mortals?

I am grappling in the Estonian MOD with the challenge of grading and assessing threats and the likelihood of dealing with them. It sounds to me if a prediction market approach might have real value for us in deciding priorities. How would you all approach this. Thanks
Glen

Its one of the techniques (and John can answer better than me!) However there are other substanital approaches that we have developed and tested within a military environment. In particular moving the high abstraction languages and also focusing risk on the attractor/barrier conditions rather than outcomes. More than be explained in a simple posting but there are some interesting development here