BRAIN

The ability to predict the outcome of future events quickly and accurately is critical in today's business environment. Agile sensing of climate changes can help us alter both spending and production without the inevitable lags inherent in the traditional marketplace. Forecasting is an integral part of the business management framework. However, month-to-month prediction is difficult in our business and data environment. Those reporting information, consciously or not, may under or over-report numbers. Some people are more risk-loving than others, and this may be reflected in forecasts. Further, individuals are aware to whom they are reporting, and that can play an important strategic role in reports. Compounding this problem, given that the current system is hierarchically organized, it takes a lot of time to pass through the layers, and distortions may be further amplified.

To address this issue, we have developed a process that takes much of the bias, manipulation, and hierarchy out of prediction. The BRAIN Process is an information aggregation tool that harnesses all of the power and truth-telling properties of market mechanisms and implements it with all of the simplicity and robustness of a simple survey. It is designed to perform in environments where market mechanisms are either not useful (due to the environment) or non-viable for complexity reasons.

To customize our mechanism, we briefly run a market game, allowing participants to buy and sell assets whose values are tied to the outcome of some future event (such as revenues, the outcome of a deal, or the success of a project or product). We monitor the buying and selling patterns of the participants, the market prices of the assets, and the true outcome. We then use that data to construct a quantitative behavioral profile of each participant, known as a beta coefficient. This beta term summarizes each participant's risk attitude, and their predictive power.

Once we have calibrated a reliable behavioral coefficient for each participant, the BRAIN Process is implemented. One could think of it as a betting game with a behavioral twist. We ask each participant to place bets (from a finite pool of betting tokens) on the possibility that various outcomes might occur. We also have a mechanism to figure out which bets are being entered based on publicly available information, and which are based on private. We then aggregate all of the participants' bets, weighting them by their individual coefficients. This aggregate, weighted prediction has been shown in laboratory experiments to outperform the market, as well as the best individual in the room.

We have implemented a pilot of the BRAIN mechanism within an HP Division. We are making monthly predictions of Revenues and Operating Profits, with the participation of a small team of senior executives.

For a deeper discussion of the BRAIN Process, please see our papers on the topic, at http://www.hpl.hp.com/research/idl/papers/future/index.html and http://www.hpl.hp.com/research/idl/papers/public/index.html. Or, contact Leslie R. Fine (leslie.fine@hp.com). You can find more information about the Information Dynamics Lab at http://www.hpl.hp.com/research/idl/.

For more information about our work in this and related areas, please see the Information Dynamics Web site, http://www.hpl.hp.com/research/idl/.