Dancing in the Dark: The Case for Chaos in an Age of <scp>AI</scp> Generated Answers

ABSTRACT One of the key aims of information search tools is to organise information and make it accessible and useful, reducing the chaos of an information deluge to ten blue links, perhaps with some context about the utility of the information they represent. Generative AI answer engines, such as ChatGPT, have gone one step further, reducing the ten blue links to a single passage of text: highly digestible, easily accessible, possibly false and ultimately bland. All chaos, including disagreement between sources, possible false information, and the influence of societal biases, and the opportunity to access more information is hidden under a rug of simple, authoritative sounding text. Friction, disagreement, reflections and odd informational juxtapositions have long been known to underpin serendipity, inspiration and creativity, yet these answer engines reduce all of these opportunities, particularly in the context of complex information needs. Using ‘bounded chaos’ as a framework, we critique the trade‐offs made by answer engines, demonstrating that, even as they are non‐deterministic, they hide the wrong parts of the chaos for those who seek learning or inspiration. We make the case against bland, and defend information seekers need for perceptible heterogeneity, friction, and even a little chaos.