• Esgin, E. & Erdoğan, Y.
• Proceedings of Third International Instructional Technologies & Teacher Education Symposium (ITTES 2015), 09-11 Sep, Trabzon.

---

Abstract
Pedagogical agents are interactive software agents that have educational purposes and human-like characters used in virtual learning environments (Esgin, 2010). This research is to adapt Pedagogical Agent Persona Instrument (API) to Turkish and to investigate the psychometric structure of it. Original form of the 25-item 5-point Likert scale was developed by Ryu and Baylor (2005) for measuring learner perception of pedagogical agent persona. They conducted exploratory and confirmatory factor analyses with three samples of undergraduate students working within the Multiple Intelligent Mentors Instructing Collaboratively (MIMIC) agent-based research environment. The final model proposed by them identified four pedagogical agent persona factors: Credible, Facilitating Learning, Engaging, and Human-like; and two latent variables: Informational Usefulness and Affective Interaction. Turkish version of the instrument is intended to be used for designing pedagogical agents in multimedia learning which are best perceived by Turkish learners. Firstly, 3 language specialists translated the 25-item API from English into Turkish independently of each other to develop Turkish form of the scale. After the integration of these translations by the researchers, 8 Turkish specialists rated it according to the “Understandability Rating Form”. All the items were above the average understandability rate of 7.5 out of 10 points. Hence, the translation accuracy was confirmed and language adaptation part of the study was completed by making suggested changes in some items. To check out the linguistic equivalency between the original and Turkish forms, both forms were implemented to 21 bilingual undergraduate students with interval of two weeks after working within the pedagogical agent-based virtual learning environment. It was found that there is a significant, positive and high-level relationship (r=0.75; p is smaller than 0.01) between the forms. After confirming equivalency between English and Turkish versions; in order to test the content validity of the scale, 9 instructional technologies experts were asked to rate each item between 0 (item does not measure pedagogical agent persona at all) and 10 (item definitely measures pedagogical agent persona) via “Measurability Rating Form”. All the items were above the average measurability rate of 7.5 out of 10 points. Therefore, the content validity of the Turkish form was confirmed. To examine the construct validity, item statistics and reliability; Turkish version of the scale was administered to 97 undergraduate students after working within the pedagogical agent-based virtual learning environment. Three items were removed from the scale and two factors were determined according to the exploratory factor and item analysis. The two-factor model explained 58.26% of the total variance of the 22 items. Reliability analysis was conducted to evaluate how the remaining items were consistent within the factors. The overall reliability of Cronbach’s Alpha coefficient was 0.94. The internal consistency of items for each factor showed high reliabilities: 0.95 and 0.86. This indicated that items within each factor of the two-factor model were very reliable. Also, the corrected item-total correlations ranged from 0.52 to 0.81. Lastly, test-retest analysis was conducted to ensure reliability in terms of temporal stability. 37 students of the same group responded the scale again three weeks later. It was found that there is a significant, positive and high-level relationship (r=0.74; p is smaller than 0.01) between the test and retest. According to the related literature and theoretical background, the first factor was labeled as “Educational Usefulness” (15-item) and the second one was labeled as “Affective Interactivity” (7-item). As conclusion, the Pedagogical Agent Persona Scale was adapted to Turkish and 2-factor 22-item form’s psychometric structure was examined. Results of the study provide empirical support for the psychometric structure of pedagogical agent persona, together with a reliable and validated scale for its assessment.