TAMIDS Workshop on Data Science/AI/ML in Education

April 24, 2019

8:45 am - 5:30 pm

Texas A&M MSC Rm 2300B

Goals: The Texas A&M Institute of Data Science (TAMIDS) held a one-day workshop on April 24, 2019 about the emerging field of using data-driven methods, artificial intelligence (AI) and machine-learning (ML) techniques in education. The goal of the workshop was to promote collaboration and engagement between Texas A&M researchers interested in advancing educational practices through state-of-the-art data-science/AI/ML techniques. Several presentations from the workshop are linked below.

Speakers: The workshop featured industry leaders from ACTNext, Udacity and Chegg, together with Texas A&M speakers from the College of Engineering, the Mays Business School, University Libraries, the Institute for Engineering Education & Innovation, the College of Education & Human Development, the College of Vet Medicine & Biomedical Sciences, and Texas A&M Qatar. The workshop was attended by over 50 faculty, researchers and students.

Data Science Workshops: TAMIDS Director Nick Duffield has identified Operational Data Science and AI in Education as initial topics for the TAMIDS Data Science Workshop series, both having great potential for growth at Texas A&M. TAMIDS is collaborating with the Texas A&M Energy Institute in organizing a workshop on Data Science in Energy that will be held in October 2019. The TAMIDS Data Science Workshop Grants Program solicits proposals to support further one-day workshops in any area of Data Science.

Workshop Organization: The organizing committee for the workshop comprised Krishna Narayanan (ECE, Chair), Othmane Bouhali (TAMUQ), Dilma Da Silva (CSE), Michael De Miranda (CEHD), Nick Duffield (ECE & TAMIDS) and Venky Shankar (Mays Business School).

Program

Introduction

(click arrow at right of title to view talk abstract)

8:45
Nick Duffield (ECE & TAMIDS) & Krishna Narayanan (ECE): Welcome and Introduction. Slides

Session 1: Machine Learning in Education – Industry Perspective

Taking high-stakes assessments such as The ACT Test requires the review of the knowledge and skills that are associated with the test. Given the broad scope of these subject domains, learners would benefit by receiving targeted, personalized recommended resources that align with their individually diagnosed area needs. We have created a Recommendations and Diagnostics (RAD) API that can be plugged into a learning and assessment system to continuously track a learner‰Ûªs practice assessment.

Increasingly fast development and update cycle of online course contents, and diverse demographics of students in each online classroom, make student performance prediction in real-time (before the course finishes) and/or on curriculum without specific historical performance data available interesting topics for both industrial research and practical needs. In this talk, we tackle the problem of real-time student performance prediction with on-going courses in a domain adaptation framework, which is a system trained on students’ labeled outcome from one set of previous coursework but is meant to be deployed on another. In particular, we introduce a GritNet architecture, and develop an unsupervised domain adaptation method to transfer a GritNet trained on a past course to a new course without any (students’ outcome) label. Our results for real Udacity students’ graduation predictions show that the GritNet not only generalizes well from one course to another across different Nanodegree programs, but enhances real-time predictions explicitly in the first few weeks when accurate predictions are most challenging.

This talk reminds participants of the promises and pitfalls when data sciences are used to focus on student learning, as well as retention and progress towards degrees, and overview some existing practices and tools for enhancing education with data science. The point is that the data scientist must not avoid the policy of use as they mine the data.

10:30
Break with Refreshments

Session 2: Data Science in Education

This talk reminds participants of the promises and pitfalls when data sciences are used to focus on student learning, as well as retention and progress towards degrees, and overview some existing practices and tools for enhancing education with data science. The point is that the data scientist must not avoid the policy of use as they mine the data.

With the emergence of commodity data science, machine learning can now be leveraged to enhance departmental programs and student progression. Available tools can be applied to visualize curricular dependencies, identify challenges within a program and hurdles on the path to degree completion. In turns, these findings can inform decision makers on where to allocate resources in an effective manner. The presentation will include a case study based on the visualization of prerequisites.

As partners in AI/ML, the Libraries would like to share our relevant interests/experience. We are most connected to Building Infrastructures and Ethics of using AI. We have:+skills toward developing DS/AI/ML projects that humanely automate problem-solving for a better quality of life
+infrastructure in place
+desire to build our capacity to support campus needs
+professional ethics that hardwire concerns for equity, privacy, IP rights, and fairness into our everyday duties. As partners in AI/ML, the Libraries would like to share our relevant interests/experience. We are most connected to Building Infrastructures and Ethics of using AI. We have:
+skills toward developing DS/AI/ML projects that humanely automate problem-solving for a better quality of life
+infrastructure in place
+desire to build our capacity to support campus needs
+professional ethics that hardwire concerns for equity, privacy, IP rights, and fairness into our everyday duties

11:55
Session Panel Discussion
12:10
Lunch (provided to registrants)

Session 3: Machine Learning in Education – Computer Science Perspective

Personalized education needs computers to understand students and adjust teaching methods accordingly. To achieve this goal, the first step is to use machine learning to understand the facial expressions and body languages of students. In this talk, I will introduce methods in deep learning for understanding students, and outline a roadmap for the next-generation education system that helps every student learn better.

Public speaking anxiety ranks as a top social phobia in the U.S., especially for minority groups and first-generation students. This research exposes students to public speaking through virtual reality interfaces, and designs machine learning models for quantifying and intervening upon public speaking anxiety. Findings suggest that virtual reality can mitigate anxiety captured through self-reports and signal-based indices, laying the foundation towards automated bio-feedback interventions.

A large portion of learning material is in text forms, including textbooks, class notes, assignments, quiz questions and answers etc.. Clearly, automatic text understanding techniques, once applied, have great potential to improve the quality and effectiveness of education. In this talk, I will briefly describe a few recent advances we made on two specific tasks, 1) word sense disambiguation and 2) discourse parsing, for achieving automatic understanding of word semantics and text structures.

Using sketching in education provides enormous potential to enhance the student learning experience. In this discussion, we will look at several projects in the Sketch Recognition Lab in which we are using data analytics paired with the power of artificial intelligence. Our goal is to produce better algorithms and smarter interactive systems which can closely resemble aspects of human interaction. Our projects include tools to help teach drawing, science, and engineering concepts.

2:50
Session Panel Discussion
3:10
Break with Refreshments

Session 4: AI in Education

A plethora of scientists have worked hard on all fronts of the artificial intelligent spectrum with limited breakthroughs in learning environments. Our web-based intelligent tutoring systems are built on carefully defined parameters and proven to be very successful in improving literacy. Some myths about AI design will be dispelled during this talk and results that have been showcased by Amazon and US News will be presented. References: (1) Wijekumar, K., Meyer, B.J.F., Lei, P., Xuejun, J., Joshi, R.M. (2017, April 5). Evidence of an Intelligent Tutoring System as a Mindtool to Promote Strategic Memory of Expository Texts and Comprehension with Children in Grades 4 and 5. Journal of Educational Computing Research.  DOI: 10.1177/0735633117696909 (2) Wijekumar, K., Meyer, B.J.F., Lei, P. (2012). Large-scale randomized controlled trial with 4th graders using intelligent tutoring of the structure strategy to improve nonfiction reading comprehension. Journal of Educational Technology Research and Development. 60, 987-1013.

AI is reshaping how we learn. AI could hugely impact education in low-income countries with 6 billion people. The school dropout rates here are astounding (42% in sub-Saharan Africa; 33% in west and South Asia). Completing secondary education can lead to a 75% rise in per capita income and accelerate poverty elimination by 10 years. The presentation will discuss research that creates and tests personalized AI based learning solutions at scale.

In several of our research grants, we, from the Center for Research and Development in Dual Language and Literacy Acquisition (CRDLLA) and Education Leadership Research Center (ELRC), have developed and tested Virtual Mentoring and Coaching across teachers. We are interested in translating this form of real-time feedback to in-service and pre-service teachers to an AI environment..

Learning analytics can be strategically utilized to support student success from enrollment to commencement. The Center for Educational Technologies in the College of Veterinary Medicine & Biomedical Sciences is leveraging the power of data to develop predictive models that evaluate course content, assess student progress, predict performance, and identify at-risk students in the Doctor of Veterinary Medicine program. Interactive dashboards will bring the data to life for each stakeholder level.

4:50
Othmane Bouhali (Advanced Scientific Computing Center, TAMUQ): A proposed platform for learning enhancement.
5:10
Session Panel Discussion

Conclusion

5:20
Nick Duffield (ECE & TAMIDS): Wrap up and next steps
5:30
Finish