For example, if a student is highly conscientious (such as ‘Peter’) then the teacher may assume that have put in considerable effort and praise them even if they achieved a bare pass. neuroticism compare ‘David’ and ‘Andrew’). conscientiousness compare example learners ‘Matt’ and ‘Peter’ in Table 1) and how they would respond to their progress (e.g. We postulate that teachers use the learner’s personality as a guide for both the effort they put in (e.g. It may depend on how much effort the learner has put in, Footnote 1 how you believe they will be feeling about the score, and how they may react to feedback. However, what if the learner only achieved 55 %? We believe that the answer to this question depends on the circumstances surrounding the learner.
Would you provide praise if the learner scored 90 %? Most likely, yes. Would you praise them? Obviously, this would depend on the score they achieved.
This paper investigates adaptation to an under-explored learner characteristic, namely personality, in particular the traits in the Five Factor Model (also known as the Big 5).Īs feedback is an important part of learning and motivation (Deci and Ryan 1980), we investigate how a conversational agent could adapt feedback to learner personality and performance.Ĭonsider the learners in Table 1 – suppose they have all achieved the same score on a test. 2014), learner skills (Desmarais and Baker 2012) and performance on a task (Varnosfadrani and Basturkmen 2009). 2008), learning styles (El-Bishouty et al. 2009), motivational state (McQuiggan et al. Common learner characteristics for this adaptation include affective state (Nkambou 2006 Woolf et al. As also reported in this journal, there has been considerable research in developing Intelligent Tutoring Systems and Adaptive Learning Environments which intelligently adapt the learning environment to learner characteristics (e.g. Keeping learner motivation high is a key challenge in digital educational systems, with the lack of personalized approaches traditionally delivered by human tutors increasing drop out rates. Finally, we ran a qualitative study with teachers to investigate the algorithm’s effectiveness and further refine the algorithm. A refined algorithm was created based on the results. Two algorithms were created using different methods to describe the adaptations and evaluated on how well they described the experimental data using DICE scores. neurotic individuals with poor grades received more emotional reflection). The type of emotional support given varied between different personalities (e.g. Five experiments were run where participants gave progress feedback and emotional support to students with different personalities and test scores. We investigate the adaptation of this feedback to a learner personality, in particular the traits in the Five Factor Model. Secondly, we investigate which emotional support messages the conversational agent should use (for example: using praise, emotional reflection, reassurance or advice) given learner personality and performance. Firstly, we investigate whether the conversational agent should employ a slant (or bias) in its feedback on particular test scores to motivate a learner with a particular personality trait more effectively (for example, using “you are slightly below expectations” versus “you are substantially below expectations” depending on learner conscientiousness).
As feedback is an important part of learning and motivation, we investigate how to adapt the feedback of a conversational agent to learner personality (as well as to learner performance, as we expect an interaction effect between personality and performance on feedback).