Search engines, cognitive biases and the man–computer interaction: a theoretical framework for empirical researches about cognitive biases in online search on health-related topics

In the present paper the term bias is used according to the definition conceptualized within the ‘heuristic and biases’ program (Tversky and Kahneman 1983). In this perspective a bias is a systematic deviation from norm or rationality and they are adaptive as result of automatic and intuitive reasoning methods (the heuristics) that simultaneously afford quick decision-making and reduce the risk of major consequences or missed opportunities.

Being aimed at discovering the way users actually cope with online information, the proposed framework will take search engines as they presently are. This does not imply that the decisions of those who actively shape the online information-retrieving systems (site owners, engineers etc.) are not to be questioned. Especially the lack of transparency of Google rankings is a consistent, overall reaching violation of any sound research criterion. In our opinion, however, a research about the cognitive styles used in the present situation is propedeutic to any proposal of reform. We should not merely wish for an abstractely fairer information-retrieve system, but for a system which is fairer for the actual people who use it, with their biases and cognitive styles. Moreover, any research which concentrate on the willingly manipulations from the active information producers cannot reach its aim if it does not take into account that those manipulations pander on existing biases. If, for example, the representative heuristic did not exist, political propagandists would have a harder time to shape opinions.

‘Informational utility’ can be defined as the usefulness a piece of information has (or is perceived to have) in order to adapt to the environment. “An individual will select a mass media message when he estimates that the message reward value exceeds the expenditures incurred in obtaining it. Instrumental utility is one major component in reward value” (Atkin 1973). Informational utility helps to explain why, in certain cases, selective exposure can work against the avoidance of attitude-dissonant messages (Knobloch et al. 2003). For example, Iyengar et al. (2008) have detected that, before the 2000 US presidential election, republican electors showed a preference for republican media, whereas democrats showed no comparable bias. Knobloch-Westerwick and Kleinman (2012) suggested that, since democrats felt that they were going to lose the elections, they found useful to expose themselves to messages from the other side, to learn about future politics, and to develop counter-arguments.

During the construction of the experimental design, it should be considered that the formulation of a hypothesis and the following search activity about the possible causes of a cluster of symptoms may be very different from a search activity about the prognosis and therapies of an already diagnosed condition; the cognitive processes involved could be very different. It is advisable to construct two separate sets of experiments, each dealing with these two kinds of activities independently.

This is an issue, for example, for Wikipedia, where sometimes sources cited in support of an article are actually referring to information taken by that very article.