Assessment of answers to questions about a text and knowledge of cognitive processes

This study examines to what extent assessment of text comprehension involves knowledge of the properties of human cognition (theory of mind) and the social context of assessment. The subjects (N=332) were asked to read a text and then assess eight answers to questions about this text. The independent variables were the quality of the answers to requests for paraphrases, the quality of answers to direct questions about the meaning of the text, the order of the paragraphs in the text and the human vs. artificial source attributed to the answers. Results show that answers to requests for paraphrases were thought to be better when the source was artificial rather than human. Inversely, answers to direct questions about the meaning of the text were thought to be better when their source was human. The assessment of answers attributed to a human source were differentiated by a greater integration of contiguous assessments (contrast effect between contiguous assessments). This was noted more particularly for a person than for a machine, poor paraphrasing being followed by a better assessment of answers to questions about the meaning of the text. The assessment of human understanding of a text is hypothesised to be guided by an expectation of answer coherency and a wider and more structured knowledge than the assessment of artificial answers.