A Critical Scholarly Conversation with Tony Bates: Insights on the Future of Online and Distance Education

On Digital Transformation and Learner Agency

Aras Bozkurt:

The term ‘digital transformation’ is now central to higher education strategy. While it promises efficiency and data-driven personalization, it can also lead to highly structured, platform-centric experiences that may inadvertently limit student choice. From your perspective, what is the real impact of this widespread transformation on learner agency? How can institutions, and more specifically instructors, design online learning environments that use technology not just to deliver content more efficiently, but to genuinely empower students to take ownership of their learning paths, make meaningful choices, and co-create their educational journey?”

Tony Bates:

The issue here is pedagogy, by which I mean different ways of teaching. To oversimplify, there is the behaviourist approach, based on content delivery, understanding and testing, and the constructivist approach, which focuses more on the development of a wider range of learning skills, such as critical thinking, analysis, and problem-solving. The distinction is not absolute. Content and understanding are a pre-requisite for skills development. However, learner agency in the past has tended to be associated with constructivist approaches although again there are exceptions.

Second, there is no isomorphism between pedagogy and technology. Most technologies can be used to teach in a variety of ways, and hence can be used to encourage or restrict learner agency. However, the applications of a particular technology may be more frequently associated with a specific way of teaching. For instance, Zoom and other video-conferencing technologies tend to be used mainly for content delivery, although they could equally be used for discussion and brainstorming. Similarly, in conventional education and particularly undergraduate teaching, content delivery tends to prevail. The question is: to what extent does the digital transformation actually change teaching methods and subsequently learner agency? I suggest that the evidence to date suggests that new technologies in general tend to be grafted on to traditional ways of teaching and learning, rather than transforming learning through increasing learner agency. For evidence of this, look at how technology was used during Covid-19 Pandemic. Most institutions continued using video-conferencing for lecture delivery.

However, recent developments in AI, in particular large language models (LLMs) such as ChatGPT, have placed powerful tools in the hands of learners, thus, theoretically increasing their agency. Students can use ChatGPT to identify what they need to learn, get feedback and assessment on their learning, and find out what courses and programs are available to them (see for instance Contact North’s AI tools for students). However, these tools focus very much on content delivery and assessment, and do not promote the development of higher order learning skills. In general, traditional institutions still control the accreditation process, which may limit learner agency, although alternative providers, such as industry-based qualifications and microcredentials, are also challenging the control of traditional educational institutions.

There are two main threats to learner agency from AI. The first is concentration of AI into a limited number of providers. Current LLMs require massive amounts of financial investment and have high overheads in the form of data entres and electricity. At the moment, such tools are freely available, but it is not clear yet how the main providers of AI intend to monetise their software and this may have major implications for learner agency. The second threat is limiting learners to the delivery and testing of content, and not enabling the development of higher-level learning skills. However, it is a matter of time before other forms of AI are developed, based on cognitive approaches, that will facilitate the development of higher-level learning skills.

On Constructive Human-AI Interaction in Pedagogy

Aras Bozkurt:

With the rapid integration of generative AI, the nature of interaction in education is fundamentally changing. Looking beyond its use for simple information retrieval or the challenges of academic integrity, what do you see as the most promising models for constructive Human-AI interaction in online education? How should a course designer or a frontline instructor rethink pedagogy to foster a relationship where AI acts as a creative partner, a critical thinking scaffold, or a personalized tutor? What new skills and literacies do both educators and students need to develop to thrive in this new interactive landscape?”

Tony Bates:

As in my previous answer, educational applications of AI technology today are based on LLMs and as such are focused on content delivery, feedback and assessment. At this they excel and provide very satisfactory human-AI interaction. Being probabilistic and hence behaviourist, however, they do not lend themselves to skills development beyond comprehension. This will require major breakthroughs in AI algorithmic design, based on cognition. Given the concentration of investment in LLMs, this may take many years, if ever.

However, LLMs require a major shift in teaching methods by instructors. They need to shift their focus from content delivery and the assessment of comprehension, which AI can do much more cost-effectively, to higher-order skills development. This will mean using other teaching methods than lectures, such as discussion, problem-solving, and original student-led research, while at the same time incorporating LLMs-based learning to provide the necessary content and background needed to support the development of such skills.

On Cultivating Learner Autonomy in the Digital Knowledge Age

Aras Bozkurt:

The concept of learner autonomy is more critical than ever in an age of information abundance. However, being an autonomous learner today seems to require more than just self-discipline; it demands sophisticated digital and critical literacy skills to navigate misinformation, evaluate sources, and synthesize knowledge. From your practical standpoint, how can online education be explicitly designed to cultivate this modern form of autonomy? What specific instructional strategies, learning activities, or assessment methods have you seen that are most effective in helping students move from being passive consumers of content to becoming active, self-regulating, and critically-minded directors of their own learning?”

Tony Bates:

High quality instructors, and in particular school teachers, have always known that students require basic or fundamental skills, such as literacy, verbal communication, collaboration, focus, and respect, that transcend any particular subject domain. These fundamental skills however are not taught in most cases separately from subject matter but are deliberately and unobtrusively embedded in the teaching-learning process. Digital literacy and learner agency are similar skills that need to be embedded within any subject domain. However, although embedded they must also be deliberately taught, clearly identified to students, and assessed. Unfortunately, most higher education instructors are not trained in such methods and are primarily experts in content. This suggests a team approach to teaching, with instructional designers guiding subject experts on ways to develop such ‘soft’ skills.

On a Pragmatic Vision for the Future of Online Education

Aras Bozkurt:

Drawing on your extensive experience advising governments and institutions globally, what is your vision for the next decade of online and digital learning? Beyond the hype cycles of new technologies, what are the most significant systemic challenges—be they pedagogical, financial, or political—that higher education must overcome to deliver on the promise of high-quality, accessible, and flexible education for all? If you were to give one core piece of advice to institutional leaders attempting to build a robust and sustainable digital learning strategy for the future, what would it be?

Tony Bates:

AI can already effectively teach most standard undergraduate content better than humans. This means subject experts need to change their focus from content delivery to developing the necessary skills that students will need to become expert in the subject area. Instructors will also need to focus on the topics and areas either not well recorded in the last few years (because LLMs have insufficient data on recent developments) and ‘outliers’, exceptions or unique cases that the LLMs’ probability algorithms miss or ‘hallucinate’. In particular, instructors will need to help students to develop digital literacy skills particularly relevant to their subject area. More emphasis needs to be put on moderated student discussion, original research, and project work that builds on the content available through AI.

Higher education also need to press government for greater regulation of AI companies, and in particular to prevent monopolies, or over-concentration in just a few companies. This means putting substantial public money into universities for AI research, to balance the efforts of commercial entities such as Microsoft.

Lastly, as well as incorporating AI and changing teaching approaches, institutions should carefully evaluate and research their selection and applications of AI, to ensure reliability, accuracy and relevance. In particular, they need to monitor and evaluate the next wave of AI development based on cognition when it becomes available. This means having at least some in-house expertise in AI and teaching.