Using Eye Movements to Develop an Ecologically-valid AI Measure of Reading Comprehension

The ability to understand written information is a critical skill for surviving and thriving in a literate society. As such, reliable and accurate measures of reading comprehension ability are necessary to ensure that readers with difficulties understanding texts can be identified and provided with adequate help as early as possible. However, in recent years the validity of reading comprehension assessments used in clinics and schools has been brought into question.This thesis examines the validity of existing measures of reading comprehension ability typically used by schools, clinics, and research. Specifically, it examines the influence of the way reading comprehension tests are designed on the skills they measure, and readers’ performance on the test. In addition, the thesis examines the possibility of using eye-movement behaviour during reading as an alternative to standard measures of reading comprehension ability. The results contribute to our understanding of reading comprehension as a cognitive skill, and suggest that reading comprehension measures should be chosen with caution as they do not always measure the same aspects of reading comprehension. The results also provide support for the potential of using eye-movement behaviour as an alternative to standard measures of reading comprehension ability. The current thesis, together with future research on the relationship between eye-movement behaviour and reading comprehension ability provide a strong basis for the development of a reading comprehension assessment based on eye-movement behaviour during reading