Advisor

Narrowing the Learning Gap with AI-Powered Educational Apps

Posted August 24, 2021 in Data Analytics & Digital Technologies
Robot AI

I have become fascinated with the application of artificial intelligence (AI) for educational purposes. The reason is that AI-powered educational applications hold the promise of helping to narrow the learning gap by providing educational opportunities to people of all ages regardless of gender, economic/social status, or the availability of established schools and other traditional educational resources.

In this Advisor, I examine the use of AI for personalized learning and enhanced online/distance learning platforms in the form of cloud-based and mobile applications. I also provide some examples of commercial AI-powered educational applications.

Usage & Benefits

AI educational applications can help students meet general education requirements and assist those seeking to advance their careers through additional professional training (e.g., in accounting, marketing, programming). They are also becoming important for assisting students at universities and other learning institutions with advanced courses (e.g., courses in medicine, physics, chemistry, computer science).

The main benefits afforded by AI educational applications include the ability to distribute educational opportunities to anyone (with an Internet connection or mobile device) regardless of geography and increased personalization of the overall digital learning experience.

Functionality

AI educational applications are intended to add support, guidance, and subject matter expertise to the digital learning experience similar to how an experienced human teacher does in the classroom. This is effected by applying machine learning (ML) and natural language processing (NLP) to analyze student behavior and the student’s level of understanding of topics and courses. Based on these analyses, the software generates personalized content for students with tailored recommendations, including multimedia and other content corresponding to the grade level, age, and subject matter interests of students, and according to their strengths and learning patterns.

NLP is also applied to generate topic-specific questions to further assess a student’s understanding of course matter. This assessment is used — in conjunction with predictive analytics — to predict a student’s test results. Based on these findings, the software generates personalized study aids intended to help improve the student’s future outcomes.

Market Trends & Developments

Although AI educational software is still in the early stages, we are seeing considerable innovation. Two trends are driving developments. First, educational software vendors have embraced the potential offered by AI for personalizing the digital learning experience. Second, the coronavirus pandemic has dramatically increased demand for online learning platforms due to schools and businesses requiring their students and employees to work from home.

The AI in education market consists of on-premise, cloud, and mobile apps. Due to faster implementation and cost considerations, cloud and mobile are the current preferred deployment methods. This was particularly true at the beginning of the pandemic when schools and businesses were scrambling to support their stay-at-home students/workers. And it is expected to remain the case for the foreseeable future. As a result, educational service providers are integrating AI technologies into their cloud platforms and are implementing mobile AI educational apps to assist students with homework and class work. Schools are also developing custom AI applications for tutoring and other educational purposes.

Walden University’s Julian

Walden University is developing an AI-powered tutor to assist student learning via its professor-conducted online courses. The application, named “Julian,” was built using Google Cloud’s AI and ML capabilities. Julian is intended to help students master concepts covered in on-demand learning activities by engaging them in a dialogue conducted via chat. It also offers personalized learning activities and evaluates student responses and provides them with feedback.

Julian supports four learning activities, including “choose the fact,” “answer a question,” “paraphrase practice,” and “knowledge notes.” These activities are auto-generated by ML language models based on Bidirectional Encoder Representations from Transformers (BERT) — neural net language models. Google was the first to introduce a transformer-based NLP model — “BERT” — in 2018 as open source. This was followed by more open source transformer-based NLP models in 2019, including Microsoft’s MT-DNN, Facebook’s RoBERTa, Google’s XLNet, and NVIDIA’s Megatron.

Julian functions as follows. Walden instructors first load the course learning content into the Julian platform and specify the required knowledge in the form of learning units. An ML component then indexes the content, linking each learning unit back to the relevant sections of the learning content. The instructor then conducts a quality assurance review to ensure the learning units and indexing have been defined correctly.

Development of Julian started in 2019 as a research project by Walden and Google Cloud. The goal was to prove the viability of building a prototype tutor app and to test it with students to gain real-time feedback based on actual student usage. (Walden University is the first higher education institution to prototype this technology with Google Cloud.)

This initial proof-of-concept app was tested by a select group of Walden's course-based sociology students and competency-based early childhood studies students. Approximately two-thirds of students in the testing group opted to try Julian and provided feedback indicating that it was “a good addition to their learning process.” According to company reps, “they found it useful in adding to their knowledge on various concepts and for completing assignments.”

The university noted that Julian’s different features became more important to students over the course of the semester. During the earlier part of the semester, students found the “knowledge notes” feature most valuable. while later in the semester, students indicated that having the tool quiz them and assess their knowledge mastery became more important. The university says it is using the students’ feedback to develop the next version of Julian, which they plan to test on a wider group of students.

Quizlet’s Learning Assistant Platform

Quizlet offers an AI-powered “Learning Assistant Platform” to enhance student study sessions on its online education service via more personalized and streamlined learning activities. Learning Assistant uses an interactive Q&A format. The user is asked a series of questions about their study goals. NLP interprets the meaning of their answers, while an ML component allows the application to analyze and adapt to a student’s study behavior over time.

Learning Assistant employs generalized ML models that have been trained on data pertaining to study insights accumulated from millions of students and how they study and progress through educational content. These domain-specific models cover biology, language, math, social studies, arts and humanities, and business subjects like computer skills. As more students use the program, more data is generated, which is then used to further train and enhance the models.

Conclusion

AI-powered educational applications hold the promise of extending educational opportunities to under-served communities and for assisting students in achieving general education requirements and those attending higher-level learning institutes.

That said, there are considerations. Primary issues with using AI educational applications center around data collection and sharing and privacy issues — issues that can be magnified when pertaining to minors. There are socialization issues, too. Long periods of online-only learning have shown that children with limited exposure to traditional, in-classroom settings may suffer from lack of in-person social interaction with peers — interactions that are currently difficult to simulate in digital-only learning environments. In the not-too-distant future, I think we can expect providers of online and distance learning platforms to apply AI in conjunction with virtual reality to impart a greater degree of social interaction capabilities within their applications, perhaps in the form of highly realistic avatars implemented using agent-based technologies. 

Finally, I’d like to get your opinion on the potential for AI in education. As always, your comments will be held in strict confidence. You can email me at experts@cutter.com or call +1 510 356-7299 with your comments.

About The Author
Curt Hall
Curt Hall is a Senior Consultant with Cutter Consortium’s Data Analytics & Digital Technologies and Business & Enterprise Architecture practices and a member of Arthur D. Little’s AMP open consulting network. He has extensive experience as an IT analyst covering technology trends, application development trends, markets, software, and services. He has extensive experience as an IT analyst covering technology and application development… Read More