Week 10: Course finish

This is the final week of the advanced HCCS course. On friday we will finish by making some links between accessiblity and usability, which should round things off well. The course has been a useful learning experience. My favourite area has definitely been multimodal technology and its application. The web accessibility aspects have been essential to me, for use in industry after my degree. Having nearly completed the accessiblity audit I can see how it may be tricky to maintain a unique personal style in developing websites, whilst keeping them fully accessible. Guidelines like WCAG 2.0 are restrictive web developers to some extent, but they are important because disabled people sahould have the same access to content as everyone else. Moreover as guidelines become more serious in the legal sphere, developers will have no choice but to comply. With luck the up and coming WCAG 2.0 guidelines will be more successful than its predecessor.

Week 9: Adaptable and adaptive systems

This week Pete and Sophie led the talks on adaptable and adaptive systems. These systems have the capacity to change and learn from experience. Adaptable systems require an input from the user to make adjustments; adaptive systems go a step further and automatically adjust by monitoring the user’s behaviour. There are many examples of adaptive software systems, particularly on the web, e.g. Amazon's personalization and the old Pandorra. The software is usually driven by artificial neural network (ANN) learning algorithms (see Sejnowski and Rosenberg’s, 1986, NETalk for an influential example). ANNs are different to traditional rule based inference algorithms. They consist of simple nodes linked in parallel interacting with higher levels through simple connection signals or weights, to produce an overall output from the resultant weight. Following a brain neural network analogous model, the computation style of ANNs is highly efficient at pattern recognition. This makes it an ideal candidate for adaptive systems because certain patterns of behaviour are picked up on and the the output is adjusted accordingly. However with highly complex networks the output behaviour is difficult to predict compared to more rule based inference algorithms. Thus it is possible that the output may turn out to be undesirable if the algorithms are poorly designed.

With technical issues aside the big debate in today’s class focused on whether adaptive systems cause people to lose control of their actions and threaten their personal identity, or whether they reduce cognitive load and free up time for creative activity. The thing is these systems are everywhere already! They are used in the economic and military world extensively, making thousands of financial decisions a second, and guiding weapons & military UAVs. They are also being used by millions of people a day over the internet in many thousands of software applications.  

The question is, how much do we want computers and software to control and guide our behaviour? I personally can see huge benefit in mass intergration of adaptive systems into our everyday lives. I would love my computer to react and adjust based on my personal use of it. For example it would be great if software prompted appropriately whilst I worked on my computer; perhaps sending me relevant research papers when it saw I was searching for a topic. The possibilities are almost endless; ordering the shopping when the fridge was low, suggesting music I might like, even perhaps reacting to my mood as research in affective computing is demonstrating is possible. Of course sometimes adaptive software gets it wrong, but that’s a limitation of the current technology, not of the principle in itself. I think ultimately it’s about finding the right balance between adaptability and adaptiveness, i.e. how much work the user does and how much the technology does. Stumbleupon I think does a good job, allowing you to input your interests and then adjusting content based on whether you rate sites. It thus strikes a balance between the user’s control and its own intelligence (if intelligence is the right word to use).

In terms of such systems benefit to AX for disabled people, they can have huge advantages. Every individual is different and adaptive systems potentially have the ability to react to individual’s needs and behaviour by recognizing and reacting on patterns. One example that was mentioned in class is ‘Leitza’, software that could help screen readers by searching ahead in the page for interesting and appropriate content. In another example the algorithms used to convert the patterns of light entering a camera into a haptic representation for blind people are often achieved using artificial neural networks. If universal usability is to become realised then I think adaptive systems are crucial.