Inferential Evidence | Andy Lowe

How do we begin to examine the role of what is basically an 18th century formula for calculating probability? That you could believe a certain point to be true or forward a reasonable proposition? And how would you assess the change of probability when new evidence was brought to light or added? The new evidence may itself be more or less likely probabilistic. It could be an estimate or it might be hard evidence. 

The Bayesian theorem establishes an inferential process; it’s about the calculation of probability under varying conditions of uncertainty. It is also used as a graphic model demonstrating visual means and using these to calculate relationships between different kinds of variables. This brings us to decision theory, and the use of tree graphs and decision analysis all of which rest upon this basic model. So the third aspect is, that it is the portrait of a reasoning process. In that sense it has a kind of logical relationship to how logic and symbolisable or mathematisable relations are being used. It is a modelling technique and a way of structuring information.

Now the reasoning process itself has a number of similarities. One is the legal definition of rationality. And this is again about making decisions under particular circumstances. In any given situation a legal subject is reasonable, if one can distinguish right from wrong and can adjust what the moral consequences of an action are and so on. This is a kind of model for rational choice and it re-appears in Bentham. You might say that Bentham’s criminological environment is defined by the idea of a calculus based upon a very simple rational choice model. And with that went his attempt to not only organise architectural environments, but to match that with a logical reform of language — a view that language would express logical relations most clearly and directly. So in a sense the architectural model of a panopticon and a linguistic model all went together to produce a model of knowledge.

All of a sudden we’re looking at something like a Bayesian syntax that runs across ideas of evidence to ways in which the gathering of evidence and the practices of police powers are also linked to managerial and one might say informational structures that are now the key aspects that transform policing. In this sense evidence ceases to become located purely within a juridical framework and is taken to a managerial organisational scale where we can begin to see relationships between military, police, management, finance. It goes on. In a sense it is a basic language. It is the thing that makes these link up in particular ways.

Because DNA has to be calculated against various kinds of variables and possible reference databases, particular kinds of gene markets and different kinds of pools, this means very directly that you cannot present DNA evidence as a single kind of categorical piece of unique source attribution. It’s always presented in terms of probabilities. This has lead to a definition of forensic evidence as universally probabilistic. All forensic evidence is probabilistic. 

Read more…