PARTE II. M´ ETODOS EN ASTROF´ISICA MOLECULAR
4.2 Soluciones al problema del transporte de radiaci´on
Experiment IV showed that participants developed only an accurate mental representation of the probability concept presented in the prediction task but not in the reaction task. These different learning processes in both tasks were reflected by eye movements, namely number of gaze shifts, fixation frequency, scanpath distance and scanpath velocity. In conclusion, mental representations acquired during the performance of the OVSST seem to be mainly based on the prediction task. It seems that both tasks require different cognitive processing, which fits into the dual-process assumption of attention: the reaction task seems to be an intuitive Type 1 process that is bottom-up controlled whereas the prediction task is rather a reflective Type 2 process that is top-down controlled and results in conscious knowledge participants are able to report.
Chapter 7: Experiment V – Learning Different Probabilities 99
7 Experiment V – Learning Different Probabilities
7.1 Introduction
In Experiment III and IV, distinct eye movement parameters reflected the learning processes during relearning and performing the prediction task as well as the lack of learning during the performance of the reaction task. Parallel to the learning curve eye movement parameters mainly decreased indicating also a reduced subjective uncertainty. However, the degree of objective uncertainty was kept constant in all previous experiments. Another point of concern mentioned before (see Introduction) addresses the question whether the degree of uncertainty influences eye movement parameters in such a way that the underlying shifts in learning processes become visible. Thus, in the last experiment of the experimental series different degrees of objective uncertainty were investigated which can be manipulated by adjusting the probability structure of the OVSST.
A basic model in decision making under uncertainty, as described earlier, is the expected utility theory proposed by Neumann and Morgenstern (1947). This theory assumes that probabilities of the outcomes are known, however, this is generally not the case in real-life. In contrast, the subjective expected utility theory (SEU) by Savage (1954) assumes that people chose the option which maximizes the subjective expected utility. However, it is also problematic to identify the subjective utility of the decision maker as humans often use fast and frugal heuristics to make their decisions. The way we perceive and evaluate the information can also influence the decision making (Gigerenzer & Goldstein, 1996). Furthermore, findings of several studies fail to find supporting evidence for the SEU theory (Slovic, Fischhoff, & Lichtenstein, 1977). Fishburn (1970) extends the SEU and also considers decision strategies and consequences of the decision maker. This last-mentioned extension of the SEU is highly relevant for the OVSST since the task also requires to develop decision strategies and reflects consequences of the decision between the three target objects. Further, during the performance of the OVSST all steps are involved in the decision process, i.e. from the identification of the decision situation to the final feedback after making the decision during the continuous processing in the working memory (see Fig. 1.2).
The OVSST, however, is not comparable with typical real-life situations, as no prior knowledge with regard to the probability distribution exists at the beginning. At first, participants have to encode the presented implicit information and develop a strategy to cope with the uncertain situation. They can only refer to prior experiences gained during the trainings session which does not included different probabilities of the OVSST. Participants acquire knowledge over time that is used for the mental model development and also influences the optimal degree of memory updating and exploration (cf. Doya,
Chapter 7: Experiment V – Learning Different Probabilities 100 2008). For instance, it is better to ignore the rare occurrence of the object reappearance at the bottom entrance, than trying to integrate this situation in the response strategy as reported in Experiment I (Chapter 3). Thus, the ignorance of irrelevant information and the focus on relevant information has to be learned and anchored in the mental representation.
In previous work, different degrees of uncertainty manipulated by probabilities were already investigated. In the experiment of Shaw and Shaw (1977) participants had to search and identify single target letters whereby some locations had a lower and a higher probability of occurrence. The authors already showed that participants were able to learn associations between target objects and high probability locations and respond more efficiently relative to low probability associations. Richer and Beatty (1987) reported in their study that reaction times increase with response uncertainty during the performance of two-choice and four choice tasks with go and no-go responses. These studies showed that different degrees of objective uncertainty influenced learning and behavioral data. Therefore, it seems to be necessary to examine in which way previous findings can also be applied to a higher degree of uncertainty. The following research questions address eye movements and judgment times related to lower and higher uncertainty and were derived from the aforementioned literature and presented data in the previous experiments.
7.1.1 Research Question and Hypotheses
In the current experiment, we compared the task performance in a high probability condition, used in the previous experiments, with task performance in a low probability condition. Two pilot studies were used to determine the threshold of lower probabilities which participants were still able to discriminate. Results of the first pilot study showed that Gabor figures with lines caused an object bias in the way that only the preferred exit of this object was learned correctly. Thus, Gabor figures were improved once again and counterbalanced across all participants to avoid biases. In the final experiment participants performed the OVSST twice, in one session they had to perform the probability concept used in the previous experiments and in another session a concept with higher uncertainty. The initial uncertainty of the participant due to the innocence of the experimental concept was reduced as participants were explicitly instructed to learn the probability concept.
Three main expectations can be proposed: First, it is supposed that that prediction accuracy is reduced in the low probability condition due to the higher degree of uncertainty. Even if participants use the optimal decision strategy, TTB, task performance should be reduced and learning times extended, because it takes longer to detect the best decision strategy since more evidence accumulation steps are necessary in order to build a realistic mental model of the probability structure. Second, it is
Chapter 7: Experiment V – Learning Different Probabilities 101 expected that judgment times in the low probability condition are longer according to Richer and Beatty (1987), because of the higher degree of uncertainty and thus different processing demands. Finally, a higher degree of objective uncertainty during decision making would probably lead to more ambiguous information and thus to a more widespread use of coping strategies. Thus, it is expected that higher uncertainty in the low probability condition is reflected by more visual search behavior to reduce uncertainty according to Lipshitz and Strauss (1997).