Principia Cybernetica Web

The Adaptive Hypertext Experiment

This experiment tries to investigate how we can make a hypertext network adapt to its users. It was set up to test a number of learning algorithms which we proposed to make the structure (links) of a hypertext web change according to the pattern of its use. The basic idea is that connections which are used frequently become easier to follow, while those which are used infrequently disappear. The experiment was run twice for a few weeks during 1995, and produced a huge amount of data. We are still analysing different aspects of those data, in order to test hypotheses about browsing, retrieval, associative structures and the application of spreading activation. For a detailed discussion of the experiment and its results, see the paper:

Bollen J. & Heylighen F. (1996): Algorithms for the self-organisation of distributed, multi-user networks. Possible application to the future World Wide Web, in: Cybernetics and Systems '96 R. Trappl (ed.), (Austrian Society for Cybernetics), p. 911-916.

To illustrate the principle, we have made the experimental, learning web again available. This is the network based on the data from the second experiment. However, it will still continue to learn, albeit more slowly, from the way you use it.


After choosing "Start" at the bottom of this page, you will enter a node from the network. It contains the following elements: a title word, that indicates the node where you are, and a list of ten words, that indicate nodes to which the present one is linked. Your task is to select from this list the word you want to jump to from the title node.

Choose the most related word. For example, if the title word is "dog", you might want to choose "cat", "pet", or "bark" from the list. If it is "car", you might choose "vehicle", "wheel" or "garage".
Be sensible: don't search for exotic or hyper-personal relationships like "My mother liked dogs, but my father hated them, he wore red socks, so I choose 'socks' ".
You should also not be overly taxonomic or hierarchical in your choices. Just choose that word that has a clear and intuitive association with the title word.

Read all ten words before you jump! If no word seems suitable, choose "More words..." at the bottom of the list, and you will be presented with a new list of ten words to choose from.

Stop when you are tired. You can jump as many times as you want. We expect you to make somewhere in between 5 and 20 jumps, though the exact number is not important. It is better to make a few focused choices, than to jump a lot and lose concentration because of tiredness.

Avoid infinite loops: if you notice you are cycling back and forth between the same words, choose a different related word from the list to start a new path.

Copyright© 1996 Principia Cybernetica - Referencing this page

J. Bollen, & F. Heylighen,

Aug 27, 1996 (modified)
Jul 25, 1996 (created)


Project Organization

Collaborative Knowledge Development

PCP Research on Intelligent Webs

Learning Webs

Prev. Next


Add comment...