Systematic Biases in Mean Estimation of Mixed Sequences with Positive and Negative Numbers
Tue-H9-Talk 6-6802
Presented by: Hannah Seidler
When we compare the best offers in two discounters or keep track of our personal best times in sports, we have to integrate information that we encounter sequentially. In this study, we investigated underlying cognitive processes leading to systematic biases in the perception and integration of sequentially presented numerical information.
Previous research has shown a tendency for participants to underestimate the mean of sequentially presented number sequences, which can be attributed to a compressed representation of numerical magnitudes (Compressed Mental Number Line, CMNL). In the first experiment, we show that the assumption of a CMNL also holds in the domain of negative numbers: As predicted by the CMNL, participants underestimated the absolute mean of sequences containing negative numbers. Interestingly, these findings only hold for purely positive or purely negative sequences. For sequences that included both positive and negative numbers the pattern reverses: Participants overestimated the mean of sequences with positive means as well as overestimated the absolute mean of sequences with negative means. In a follow-up experiment, we tested possible explanations for this effect. Specifically, we investigated memory biases in mixed sequences compared to only positive sequences using a memory recall task in addition to the mean estimation task. We discuss how the appearance of a second category (negative and positive numbers) affects the underlying processes of number integration compared to purely positive sequences.
Previous research has shown a tendency for participants to underestimate the mean of sequentially presented number sequences, which can be attributed to a compressed representation of numerical magnitudes (Compressed Mental Number Line, CMNL). In the first experiment, we show that the assumption of a CMNL also holds in the domain of negative numbers: As predicted by the CMNL, participants underestimated the absolute mean of sequences containing negative numbers. Interestingly, these findings only hold for purely positive or purely negative sequences. For sequences that included both positive and negative numbers the pattern reverses: Participants overestimated the mean of sequences with positive means as well as overestimated the absolute mean of sequences with negative means. In a follow-up experiment, we tested possible explanations for this effect. Specifically, we investigated memory biases in mixed sequences compared to only positive sequences using a memory recall task in addition to the mean estimation task. We discuss how the appearance of a second category (negative and positive numbers) affects the underlying processes of number integration compared to purely positive sequences.
Keywords: decision-making, Decision-from-Experience, number perception, negative numbers, mean estimation