If I may, I will answer in English to be more specific.
From a data mining perspective, it's inherently important that we base what we do on adequate risk-based assessments that are firmly grounded in evidence. There is too much investigation that happens on someone's best experience or best hunch, and there's too little in terms of the actual risk models.
As Mr. Brison pointed out, there's always the risk of being perceived as targeting ethnic communities. What we actually need are proper data-driven models to demonstrate that we are not targeting any one person because of religion or background or area of travel, but because we have an effective risk-based model.
I'd be happy to share with you some of the publications that we've done, including a publication that's about to come out on risk-based modelling within Correctional Services Canada, with a very large data set within Correctional Services.
Part of this is a challenge of not just risk-based modelling, but that we have some agencies that are more research-based than others, so the research cultures, and being able to have departments and agencies that are then driven by that research, rather than inductively by people's investigative hunches.
There's a whole institutional culture issue that's separate from actually being able to develop the adequate proper models. We don't, for instance, have a very good way of currently sharing data with researchers, something which the U.K. and the U.S. have done much better.
This is not to say that government analysts aren't good at what they do, but inherently, there are some things we can do in the academic community with some of our algorithms and methods that are perhaps a bit more advanced.