Identifying Complex Patterns in Online Information Retrieval Processes

Zusammenfassung

We define a computable benchmark framework that replicates the online information retrieval behavior of users as they proceed through the alternatives ranked by a search engine. Through their search processes, decision makers (DMs) must evaluate the characteristics defining the alternatives while aiming to observe a predetermined number satisfying their subjective preferences. A larger number of predetermined alternatives requires more complex information assimilation capacities on the side of DMs. Similarly, the complexity of the algorithms defined to formalize the subsequent retrieval behavior increases in the number of alternatives considered. The set of algorithms delivers two different strings of data, the pages clicked by the DMs and a numerical representation of each evaluation determining their retrieval behavior. We illustrate how, even when providing an Artificial Neural Network with both strings of data, the model faces considerable problems categorizing DMs correctly as their information assimilation capacities are enhanced.