Despite this being a planning blog two of the top four most popular posts are not really about planning. The expert system posts have picked up a long-term steady trickle of traffic, and I suspect this is because IS students all over the world are periodically set assignments on ‘expert systems’, and the first thing they do is to follow a search engine. This, final and slightly belated post, finishes off my journey that started with the changes to some of the permitted development rules back in August 2008.
To begin at the end, you can see the expert system in action by following the links to the two demo systems that are presently online at the pas website. The decision tree is fed by the xml you can see periodically issued in our discussion forum.
Has it worked out ? Well, up to a point. There were always going to be two parts to the expert system. The first is the tree of decisions itself – this provides the rationale for concluding that a development is permitted or not. The second is the set of knowledge that goes alongside – to enable the user to understand the questions, how they relate to their proposal, and therefore answer the question correctly. I completely misunderstood how difficult it would be to arrive at a sector consensus on some of the basic interpretations – and so the second half has languished.
xforms and xml
It’s also been an interesting technological episode. I had high hopes of xforms, as being a way of joining together the content and logic (and storage also !). However, I ended up falling out of love. Almost entirely this will be down to my advanced years and lack of brainpower. I gave it a good crack over two or three days, and ended up with not much more than a headache to show for it. Perhaps (and this is only an opinion) the xform model is as big a step as moving from procedural languages to object orientated models. It isn’t just “xml with extra stuff”, it is a paradigm that you need to “get”. I tried jumping in, learning-by-doing style hoping to find tutorials / tools that would let me at least model a small part of the form reasonably quickly. I failed.
The result of this is that whilst along the way I used a spreadsheet to help generate the xml, I’ve ended up with an 80Kb dollop of code that I hand-hack with bluefish. While I’ve written a small snippet of python to validate, parse and check the tree it is a pain to maintain. It can be done, but a series of mistakes have taught me to ensure I’m in the right kind of mood before starting.
The first few “real” installations of the expert system are starting to appear on council websites. The situation with the supporting materials will be resolved soon, probably with the involvement of the planning portal.
Perhaps most importantly, I have learnt that the rigour of the expert system modelling is a very good discipline to help make good regulations. Its my view that the changes proposed to commercial development should go through consultation in expert system form – it will hopefully allow people to better empathise with normal people. Rather than eyes glazing over trying to read planning regulations, it can be done in little bite-sized pieces using the challenge “How would a normal person respond to this question ?”. Planning regulations for normal people ? Well, I can dream.