Levari Exp.20: 2D - Partial Feedback

Overview

Suppose that observers are trying to classify a spot on the skin as normal or abnormal and suppose that the two attributes that are important are the color and shape of the spot. The investigators have found that perceptual decisions of this short are shaped by the prevalence of the target abnormality and by the feedback that observers (Os) receive. If abnormal spots are rare (low prevalence), Os will tend to become more conservative about calling spots abnormal. In this experiment, Os see items defined by color and shape. They are looking for one combination (bumpy green). Bumpy green targets can be common (50% prevalence) or rare (10%). Os in one group will get feedback about their responses based on color. The other group will receive feedback based on shape. The investigators will look for effects of prevalence and of the type of feedback. The goal is to better understand perceptual decisions in settings like clinical evaluation of skin lesions.

Suppose that observers are trying to classify a spot on the skin as normal or abnormal and suppose that the two attributes that are important are the color and shape of the spot. The investigators have found that perceptual decisions of this short are shaped by the prevalence of the target abnormality and by the feedback that observers (Os) receive. If abnormal spots are rare (low prevalence), Os will tend to become more conservative about calling spots abnormal. In this experiment, Os see items defined by color and shape. Items vary in dominant color from purple to green. Items vary in shape from relatively smooth irregular blobs (nicknamed "Bouba") to relatively bumpy blobs (nicknamed "Kiki"). Os are asked to find "Green Kikis". Since color and shape vary continuously, the border between Green Kiki and not Green Kiki will not be clear to the Os. Os will be asked to label each item as Green Kiki or not. Saying "maybe" is not an option. The between-subject variable is the type of feedback Os receive. One set of Os will get feedback about color. That is, if the observer says "yes, this is a green kiki", they will be told only "Yes, this was a green item" or "no this was not green". The other group will get shape feedback: e.g. "Yes, this was a Kiki.". The within-subjects variable is the target prevalence. The investigators can divide the stimuli into four quadrants: Green Kiki targets, and Green Bouba, Purple Bouba, and Purple Kiki distractors. In the high prevalence trials, 50% of the stimuli will be drawn from target quadrant (green kiki) and 17% from each of the remaining three quadrants. In the low prevalence trials, 10% of the stimuli will be drawn from the target quadrant while 30% will be drawn from each of the other three quadrants. The whole space of stimuli has 10 shape values and 10 color values for a total of 100 possible types of stimulus. The key dependent is the proportion of target-present responses for each type of stimulus. To analyze the data, the investigators will create a heatmap showing the percentage of times Os judge each of the 100 categories (10 colors X 10 shapes) as the target, separately for the high prevalence and low prevalence block. The investigators will calculate the response difference between the two blocks within subjects and will calculate the difference between shape and color feedback between observers. Observers with an average accuracy below 70% will be excluded. Prior experiments have a conservative effect size of \~0.8 for differences between high and low prevalence. Some effects in experiments proposed here may have more subtle effects so power will be calculated based on an effect size of 0.6. To detect such a difference between, for instance, the 50% points on two functions like those in the data figures above, requires 26 observers to achieve alpha of 0.05 and power of 0.9. The investigators will aim for 26 observers in each condition.