When I think of personalization I mainly think of systems like the old style portals, or maybe the new generation of start pages like netvibes or igoogle. Sites where to personalize the user has to build their page using whatever tools are provided. Often these pages, despite efforts at easy design, are complicated or perceived as complicated. Often perception of complexity is worse since at least theoretically a complex design may be simplified.

When I talk about ambient personalization though I am thinking about the construction of intelligent systems which learn from how the user uses the website and then go on to improve the experience for that user. Simple versions of this are already quite common. Amazon.com, for example, uses the items that a user views to help build recommendations for the user. The power of even this relatively simple example can be clearly seen in the pages of my credit card statement.

My presentation linked below is an attempt to explore some of the possibilities and some of the pitfalls inherent in asking a system to try to learn to understand users.

ambient personalisation (PDF 2.4Mb)