Carnegie Mellon University

Peer-Reviewed Publications

Publications about RoboTutor

Saxena, M., Pillai, R., & Mostow, J. (2018). Relating Children's Automatically Detected Facial Expressions to Their Behavior in RoboTutor (Student Abstract). Proceedings of the AAAI Conference on Artificial Intelligence32(1).

Goswami, M., Mian, S., and Mostow, J. (2019).  What’s Most Broken? A Tool to Assist Data-Driven Iterative Improvement of an Intelligent Tutoring System.  Proceedings of the AAAI Conference on Artificial Intelligence, 33(1).

Mian, S., Goswami, M., and Mostow, J. (2019). What’s Most Broken? Design and Evaluation of a Tool to Guide Improvement of an Intelligent Tutor. In: Isotani, S., Millán, E., Ogan, A., Hastings, P., McLaren, B., Luckin, R. (eds) Artificial Intelligence in Education. Lecture Notes in Computer Science, vol 11625. Springer, Cham.

Agarwal, M., and Mostow, J. (2020).Semi-Supervised Learning to Perceive Children's Affective States in a Tablet Tutor.  Proceedings of the AAAI Conference on Artificial Intelligence 34(9).

McReynolds, A. A., Naderzad, S. P., Goswami, M., and Mostow, J. (2020).  Toward Learning at Scale in Developing Countries: Lessons from the Global Learning XPRIZE Field Study.  Proceedings of 7th ACM Conference on Learning @ Scale (L@S '20), pp. 175 - 183, August 12–14, 2020, Virtual Event, USA.

Boote, B., Agarwal, M., and Mostow, J. (2021).  Early Prediction of Children's Task Completion in a Tablet Tutor using Visual Features (Student Abstract). Proceedings of the AAAI Conference on Artificial Intelligence35(18).

Types of Project LISTEN research studies

Test overall effects by comparing pre- to post-test gains to control treatment(s)

  • Reading Tutor rivaled human tutors in some skills (JECR03)
  • Reading Tutor beat Sustained Silent Reading (JECR07, JECR13)
  • Reading Tutor helped ELLs in Canada (IDEC09), Ghana (ITID10), India (Dev10)

Assess student skills, validating on performance in tutor and on gold standard tests

  • Cloze questions predict vocabulary and comprehension (TICL04)
  • Prosody predicts proficiency (CALICO04, TICL04, TSLP11)
  • LR-DBN subskill tracing cuts prediction error in half  (*EDM12)

Identify important variables using correlational analyses  

  • Gains decrease with time spent picking stories (ITS02)
  • Labs average higher usage than classrooms (Scale-up 07)

Compare types of practice using growth curve micro-analyses

  • Wide reading builds fluency faster than rereading (SSSR05, ITS08)
  • Students gained more from stories they picked (AIED07)

Evaluate tutor interventions using randomized within-subject experiments

  • Test effects of vocabulary intervention on taught vs. untaught words (ETS02)
  • Compare effects of tutor help on next encounter of word (SSSR04, ICALL04, *AIED13)

Mine tutor logs to discover useful knowledge

  • Identify prosodic indicators of reading gains (*FLAIRS12)

* Best Paper or Nominee

See Project LISTEN's 100+ publications online.