Effect to cause In finance, with the analyst's models of how causes lead to and are indicated by effects in a market, our system can perform abductive inference to guess causes from effects. Observations of effects in markets and economic indices are explained in terms of the hidden causes that are the best hypothesis in terms of plausibility, simplicity, closeness of match, and other factors. Aetion's approach to this type of reasoning is so powerful and general that it has been successfully applied in domains from medicine to industrial process control, and we are now working with government customers on intelligence analysis applications.

Model uncertainty It may not be clear which financial models best apply to a particular situation. Often, to varying degrees, several models will. Our system can regard the models themselves as part of the hypothesis, using the evidence to guess which parts of which models apply, and using those to test its predictions.

Preparation Often, one can anticipate that an event may occur: a release of new economic data, an SEC filing, an IPO, an embargo, a product release, a bankruptcy, or a Federal Reserve announcement. In advance of the event, the analyst has time to consider the alternative outcomes for that event, how the observers may react to each of those outcomes, and how those reactions could affect the markets accordingly. Our system can be primed with those models so that it may look out for effects in the markets that fit any models' predictions, reporting to the analyst that their anticipated scenario appears to be matching reality, thus quickly revealing the market's reaction to some event.

Exploiting findings Once good explanations are generated through our computer-assisted reasoning process, the models can be used to generate further predictions by which the hypotheses can be tested. Once confirmed, they can be exploited as predictors of further market activity, as positions are bought or sold short in markets where prices do not reflect common knowledge of the hidden causes behind the market effects.

Rapid response It often does not take long for the investors responsible for most of the trading volume to have reached similar conclusions about observed market activity. The value of our abductive reasoning approach is that the stream of market data may be processed automatically to generate and evaluate hypotheses, using corroboration, contradiction and causal models to identify the best explanation. The system can justify its conclusions in terms of the evidence and, to the extent that the models allow, can derive further predictions from its explanation of the current behavior of the markets. So, with the computer assistance that we offer, the analyst can be quickly provided with a convincing explanation of market activity, giving them greater opportunity to exploit this knowledge before the market adjusts.