The field of artificial intelligence (AI) is progressing rapidly and, as a data-driven technology, AI-powered tools lend themselves to a wide range of applications. In this blog we will look at potential opportunities for AI integration into teaching and learning, current case studies of successful use, and the explicit limitations and vulnerabilities of using this technology-driven approach.
Using artificial intelligence
For assessment, AI-assisted marking and feedback software goes beyond what has been achieved through multiple-choice quizzes. In STEM subjects, AI-powered tools can help significantly in the semi-automation of marking. Rather than grading hundreds of calculus submissions, an AI tool is very fit for purpose, using a technology called ‘replay grading’ ensuring that similar submissions are not marked twice, and grouping similar answers so they can be given a consistent grade at the same time. Examples of software designed for this use are Gradescope, Graide or the Mobius platform, which can automatically generate unique questions and answers (limiting academic misconduct) and these programs can even read handwritten equations using optical character recognition.
AI is not just restricted to mathematical or engineering subjects but can also be applied to speech and written text. Speech recognition software which is AI-enhanced can be used to improve captioning on videos or live feeds which means we can provide more accessible content. Machine learning software such as Grammarly is progressing to provide academic writing support and assisted marking. Tools within machine learning platforms such as TensorFlow analyses text for toxic behaviours and therefore could be applied on discussion forums or the like as a moderation tool.
For employability, AI could be used in conjunction with virtual reality to improve students’ soft skills and help to embed employability into the curriculum. Bodyswaps is a platform that provides immersive simulations such as job interview training, using face emotion and body movement detection to help analyse performance. In one study using VR with AI analytics, 84% of students reported feeling better prepared in advance of their upcoming interviews (Bodyswaps, 2021).
Personally, one of the most exciting ways to use AI and machine learning is through adaptive learning platforms. Virtual Learning Environments produce a huge amount of usage data daily, and the increased use of digital technologies in the last few years has furthered this. But often this data is overlooked and underutilised. There is the potential for AI models to provide valuable insights that can support learners’ development, through a fine-grained analysis, for example, providing autonomous learning recommendations, individualised instruction and personalised learning at scale.
There are a growing number of successful use cases for AI integration within the higher education sector and some examples are listed below:
- The Open University trialled a digital assistant (Taylor) to support disabled students and improve the student experience (Lister et al. 2021).
- The University of West England, Bristol uses the AI-powered CV360 tool to help students build their CV and receive feedback.
- Turnitin uses AI-powered language comprehension to assess subjective written work and a revision assistant that provides academic writing support to students.
- The IBM Research and Rensselaer Polytechnic Institute combined immersive technologies with an AI-powered assistant to help students practice speaking Mandarin as if they were sitting in a restaurant or garden in China.
- Staffordshire University uses Beacon, an AI teaching assistant or ‘student coach’ that recommends reading resources and connects students with personal tutors.
- Petroc college uses Century AI as an adaptive learning platform which offers bespoke learning to aid student outcomes.
Although these case studies show successful use, AI is still an emerging technology, most educational institutions are at the start of the AI maturity model (below), looking at the experimentation and exploration phase. Through targeted pilots, the use of AI within educational institutions in the next five years has the potential to progress from operational, to embedded and transformational.
Figure: Maturity model for the application of AI programs within education (Webb, 2022, 4).
Challenges and Limitations
So, what challenges and limitations do we need to be aware of? How can we benefit optimally from using artificial intelligence whilst also protecting against the risks this technology presents?
AI will only be good as the programs and software it can be situated in, if back-end process systems are quite manual then it will be more time-consuming to integrate. Data is also the critical foundation for this kind of change – AI tools will only be as good as the data they’re fed. In relation to this, using sensitive data related to individual learners requires university systems to have technical robustness and effective data governance. It is important to be transparent with chosen ways of data gathering and processing and acknowledge anxieties about the ways that data, algorithms and machine intelligence are used in education. Further considerations include the need to develop an ethical framework for using AI (Cormack, 2021). This includes collaborating with current organisations such as the Institute for Ethical AI in education (IEAIED, 2021) and utilising the Jisc code of practice for analytics (Sclater and Bailey, 2018). We must enable equality and diversity and acknowledge algorithmic biases in AI systems whereby certain groups of learners may be unfairly discriminated against.
Finally, learners may become over-reliant on AI systems, and this could reduce independent thought in their studies, or there is the risk that education could seem to become depersonalised.
In 2019 almost 3 billion pounds was invested globally in AI edtech start-ups, with the aim that AI could help transform students’ education outcomes, providing individualised learning. Since the pandemic, technology plays a greater role in the delivery of education with increased digitisation, leading to a larger output of data that can be utilised to help with sector challenges, improving higher education for teachers and learners alike. However, the limitations show that it’s important that this technology should be used as a support or an aid, rather than a replacement for existing methods.