Monday, November 13, 2006

Assignment 9: I shouldn't be a graphics designer.

For this assignment I did the perceptual analysis test. I was shown an image of a plastic water bottle, and asked to describe it to someone who couldn’t see the image, to try to give them the best understanding I could of what the image looked like. My description was then analyzed using LIWC, and it was used to make a (semi-accurate) prediction of my personality.

This process is known as Meaning Extraction. Before this study was done, a large number of people were given this task, and their text was analyzed. The researchers found 5 different clusters of words, and these 5 clusters are what they look for when someone now performs this task. The 5 dimensions were: Words on the label: Verbal thinking, Colors and text: Visual sensitivity, Bottle contents: Functional thinking, The bottle itself: Tactile sensitivity, and Light and shadow: Contextual thinking. So when someone now performs this test, they search through the text to find words that fit into each of these 5 categories.

Your results are then compared the average responses in each category, and then they gave a brief “prediction” of your personality based on how you compared to the average. Because of a high score on contextual thinking, they predicted “This is healthy in the sense that it suggests you can stand back and look at objects in a broader perspective.” From a low score on tactile sensitivity they predicted “Extending this reasoning a bit, you don't automatically imagine how an object feels. Touch is not your dominant sense. I just can't see you as a sculptor, a painter of still lifes, or a postmodern architect.” As the researchers said themselves, this part of the analysis was kind of “made up.”

The importance of this analysis, however, isn’t in predicting that “Graphic design may not be a wise career choice,” but in being able to analyze the text of a message. The researchers were able to accurately categorize my text into different categories, and correctly noted which categories I did tend to use more frequently than others. Now, none of these characteristics (to my knowledge) are related to deception detection, so this specific analysis probably would not be helpful for someone trying to find deceptive messages.

This method of analysis, however, is done in the style of many successful deception detection techniques. Zhou et al notes that when doing a linguistic analysis, accuracy rates increase greatly when only known important cues are studied, and the irrelevant cues are disregarded. I believe that the researchers followed this in their analysis, by looking only at words in certain groupings or styles to figure out which category the text belonged in. Also, they only looked for the relevant words that they noticed in their pilot study, and therefore in the current analysis the results are not muddled by other words which may or may not be relevant to the analysis. I believe that the researchers would also agree with Keila et al that different types of situations (like deception) leave a linguistic signature. Deception obviously is not studied in this analysis, but the researchers do imply that through linguistic analysis you are able to tell certain things about the person or the situation. I believe that this type of analysis is very effective, and could be adapted to deception if specific words that are identified as being related to deception are searched for.

0 Comments:

Post a Comment

<< Home