Predicting Instructor’s Intervention in MOOC forums

Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, pages 1501–1511, Baltimore, Maryland, USA, June 23-25 2014. c⃝2014 Association for Computational Linguistics Predicting Instructor’s Intervention in MOOC forums Snigdha Chaturvedi Dan Goldwasser Department of Computer Science,Hal Daum´e III University of Maryland, College Park, Maryland {snigdhac, goldwas1, hal}@umiacs.umd.edu Abstract Instructor intervention in student discus- sion forums is a vital component in Massive Open Online Courses (MOOCs), where personalized interaction is limited. This paper introduces the problem of pre- dicting instructor interventions in MOOC forums. We propose several prediction models designed to capture unique aspects of MOOCs, combining course informa- tion, forum structure and posts content. Our models abstract contents of individ- ual posts of threads using latent categories, learned jointly with the binary interven- tion prediction problem. Experiments over data from two Coursera MOOCs demon- strate that incorporating the structure of threads into the learning problem leads to better predictive performance. 1 Introduction Ubiquitous computing and easy access to high bandwidth internet have reshaped the modus operandi in distance education towards Massive Open Online Courses (MOOCs). Courses offered by ventures such as Coursera and Udacity now im- part inexpensive and high-quality education from field-experts to thousands of learners across geo- graphic and cultural barriers. Even as the MOOC model shows exciting pos- sibilities, it presents a multitude of challenges that must first be negotiated to completely realize its potential. MOOCs platforms have been especially criticized on grounds of lacking a personalized educational experience (Edmundson, 2012). Un- like traditional classrooms, the predominant mode of interaction between students and instructors in MOOCs is via online discussion forums. Ideally, forum discussions can help make up for the lack of direct interaction, by enabling students to ask questions and clarify doubts. However, due to huge class sizes, even during the short duration of a course, MOOCs witness a very large number of threads on these forums. Owing to extremely skewed ratios of students to instructional staff, it can be prohibitively time-consuming for the in- structional staff to manually follow all threads of a forum. Hence there is a pressing need for automat- ically curating the discussions for the instructors. In this paper, we focus on identifying situa- tions in which instructor (used interchangeably with “instructional staff” in this paper) interven- tion is warranted. Using existing forum posts and interactions, we frame this as a binary prediction problem of identifying instructor’s intervention in forum threads. Our initial analysis revealed that instructors usually intervene on threads discussing students’ issues close to a quiz or exam. They also take interest in grading issues and logistics problems. There are multiple cues specific to the MOOC setting, which when combined with the rich lexical information present in the forums, can yield useful predictive models. Analyzing forum-postings contents and bring- ing the most pertinent content to the instructor’s attention would help instructors receive timely feedback and design interventions as needed. From the students’ perspective, the problem is ev- ident from an examination of existing forum con- tent, indicating …

Content truncated to 500 words