Recommender in AI-enhanced Learning: An Assessment from the Perspective of Instructional Design

Recommender in AI-enhanced Learning: An Assessment from the Perspective of Instructional Design

Table of Contents




Abstract

As tools for AI-enhanced human learning, recommender systems support learners in finding materials and sequencing learning paths. The paper explores how these recommenders improve the learning experience from a perspective of instructional design. It analyzes mechanisms underlying current recommender systems, and it derives concrete examples of how they operate: Recommenders are either expert-, criteria-, behavior-, or profile-based or rely on social comparisons. To verify this classification of five different mechanisms, we analyze a set of current publications on recommenders and find all the identified mechanisms with profile-based approaches as the most common. Social recommenders, though highly attractive in other sectors, reveal some drawbacks in the context of learning. In comparison, expert-based recommendations are easy to implement and often stand out as simple but effective ways for suggesting learning materials and learning paths to learners. They can be combined with other approaches based on social comparisons and individual profiles. The paper points out challenges in studying recommenders for learning and provides suggestions for future research.

Keywords

recommendation, instruction, design; learning paths, learning resources, artificial intelligence

Introduction

Recommender systems for human learning based on artificial intelligence (AI) are a trending topic in research on Educational Technology (Roll & Wylie, 2016). Recommenders are already of great importance in various contexts, and they are the foundation of several online services (Kantor, Ricci, Rokach, & Shapira, 2010). Research on recommender systems in education has most often been approached from the view of computer science (cf. Manouselis, Drachsler, Vuorikari, Hummel, & Koper, 2011). Therefore, the paper addresses the question of how these developments can be related to pedagogical threads of discussion and how they conform with evidence from the state of research on instructional design.

In the following section, the discussion about learning paths is first rooted in the pedagogical discourse. In this context, the sequencing of instruction is seen as an essential condition for successful learning and as an important professional routine of trained teachers. The topic has motivated many theoretical concepts, practical models, and empirical research to identify sequences that reliably support learning and to detect moderating variables that enable teachers to adapt their strategies of teaching to, e.g., learning objectives or learner characteristics. Against the backdrop of this discussion, the paper then analyzes the mechanisms of current recommenders for learning and how they try to support learners.

Future research

Recommenders in AI-enhanced learning should rely on approved and tested models for learning (either from empirical studies or from machine learning). In some cases, however, it seems that complex formalizations try to overcome the weaknesses of the underlying models. Some recommenders are based on sophisticated formalizations with several elements that would need a complex evaluation scheme: “In the context of a learning style based on an Interpretive Structural Model (ISM), an adaptive learning path recommendation system is proposed comprising: (a) Fuzzy Delphi Method, (b) Fuzzy ISM and (c) Kelly Repertory Grid Technology” (Su, 2017). The empirical evidence of such models often relies on a study that compares one group of learners receiving instruction with recommendations and one control group without recommendations. If users deem the treatment favorable or if the treatment achieves higher learning outcomes, the quality of the recommender seems to be verified. However, with such a study, it is not possible to assess if and which of the recommendations have been meaningful and are superior to randomly selected suggestions: Instead of providing a control group ‘without recommendation’, it would be more appropriate to compare a control group with randomly assigned recommendations.

In double-blind clinical trials, control groups are given a placebo and compared against the application of a drug. For a personal learning environment, Chatti et al. (2013) compared 16 different algorithms: They found that “the quality of user experience does not correlate with high-recommendation accuracy measured by statistical methods”, thus, demonstrating the difference between user experience and algorithmic logic and proving the importance of comparing different recommendations against randomly assigned recommendations. From a learning science perspective, most of the recommenders remain a black box: The instructional logic remains hidden for learners and teachers alike. They are not able to comprehend how recommendations come about. With a search engine like Google, the underlying algorithms remain a trade secret. For an educational context, however, the algorithms applied should be transparent and modifiable by learners. Particularly complex mechanisms, on the other hand, delimit reflections and adjustments by users. Still more fundamentally, it might even be questioned if recommenders that are not based on sound evidence can be justified ethically.

In a learning context, recommenders follow – at least implicitly – instructional approaches of guided discovery learning, which provide structuring elements in an open learning environment. This approach is backed strongly by empirical evidence (cf. Tobias & Duffy, 2009); it ensures that learning is oriented towards achieving a learning objective without falling back to narrow sequencing strategies from behavioristic approaches which might impede the learning experience. Nevertheless, such approaches have to deal with the repeatedly proven finding that learners tend to ignore or even reject help systems, recommendations, or other advice whilst learning (e.g., Clarebout & Elen, 2009). Therefore, ensuring acceptance of such guidance and support systems is of topmost importance when designing (and studying) a recommendation system. Before designing a recommendation system, it is also necessary to consider what kind of recommendation learners expect, how learners search for learning resources, and how they construct their learning journey in a given context. Technically complex solutions are not necessarily the first choice. A thorough analysis of the target group and further instructional parameters need to be specified to identify which recommendation mechanism to choose.

Recommenders for movies or real estate are based on preference structures. Eventually, the aim is a purchasing decision and here, similarities prove to be good predictors. There are some differences between the single act of purchasing and the ongoing process of actively engaging in learning and educating. Therefore, the question remains how to conceptualize guidance in learning: Should recommendations provide proven routes that have been undertaken by others and that have been successful in the past? Or do we understand education as an opportunity for opening new horizons beyond established routes of thinking and for providing new experiences that might irritate us (Forneck & Springer, 2005; cf. Stojanov, 2012)? Education cannot only be limited to the training of common knowledge and skills, continuing to follow paths of learning from the past. Such a view on education would be cautious to present familiar resources but would choose to deliver surprising paths that confront learners with unfamiliar concepts and views. A final note: Recommender in AI-enhanced learning promise choice. By offering guidance and supporting the learner these systems try to improve the individual’s learning process. But at the same time, recommenders might contribute to eventually weakening the individual’s autonomy and self-regulation because of a dependency on external regulation. When designing recommenders, we should therefore consider deliberately how external guidance can be provided while still keeping the learners’ independence and self-regulation as the major learning objective – even in AI-enhanced learning.

About KSRA

The Kavian Scientific Research Association (KSRA) is a non-profit research organization to provide research / educational services in December 2013. The members of the community had formed a virtual group on the Viber social network. The core of the Kavian Scientific Association was formed with these members as founders. These individuals, led by Professor Siavosh Kaviani, decided to launch a scientific / research association with an emphasis on education.

KSRA research association, as a non-profit research firm, is committed to providing research services in the field of knowledge. The main beneficiaries of this association are public or private knowledge-based companies, students, researchers, researchers, professors, universities, and industrial and semi-industrial centers around the world.

Our main services Based on Education for all Spectrum people in the world. We want to make an integration between researches and educations. We believe education is the main right of Human beings. So our services should be concentrated on inclusive education.

The KSRA team partners with local under-served communities around the world to improve the access to and quality of knowledge based on education, amplify and augment learning programs where they exist, and create new opportunities for e-learning where traditional education systems are lacking or non-existent.

FULL Paper PDF file:

Recommender in AI-enhanced Learning: An Assessment from the Perspective of Instructional Design

Bibliography

author

Michael Kerres, Katja Buntins

Year

2020

Title

Recommender in AI-enhanced Learning: An Assessment from the Perspective of Instructional Design

Publish in

De Gruyter, Open Education Studies | Volume 2: Issue 1

DOI

PDF reference and original file: Click here

 

Website | + posts

Nasim Gazerani was born in 1983 in Arak. She holds a Master's degree in Software Engineering from UM University of Malaysia.

Website | + posts

Professor Siavosh Kaviani was born in 1961 in Tehran. He had a professorship. He holds a Ph.D. in Software Engineering from the QL University of Software Development Methodology and an honorary Ph.D. from the University of Chelsea.

+ posts

Somayeh Nosrati was born in 1982 in Tehran. She holds a Master's degree in artificial intelligence from Khatam University of Tehran.

%d bloggers like this: