Solving Year 1 programming problems with AI assistants (Michel Wermelinger, OU UK)

Although OpenAI’s ChatGPT is getting all the headlines in the press, there are more serious threats to the teaching of programming to novices: large language models (LLMs) that are not only specifically trained on code, like OpenAI’s Codex, but are part of programming environments, thus lowering the barrier to their use. The best-known AI coding assistant is GitHub Copilot, based on Codex. Given a prompt in plain English, it generates readable and often correct code. Contrary to Codex, which is paid, Copilot is freely available to students, which raises pedagogic and academic integrity concerns. Educators need to know what Copilot is capable of, in order to adapt their teaching to AI-powered programming assistants.

Previous research evaluated Codex quantitatively, e.g. how many problems have at least one correct suggestion (i.e. that passes all tests). Here I evaluate Copilot instead, to see if and how it differs from Codex, and look qualitatively at the generated suggestions, to understand the limitations of Copilot. I also report on the experience of using Copilot for other activities asked of students in programming courses: explaining code, generating tests and fixing bugs. The paper (https://oro.open.ac.uk/86438) concludes with a discussion of the implications of the observed capabilities for the teaching of programming.

iSpot & AI: Integrating FASTCAT-Cloud and PI@ntNET-API in the Cos4Cloud framework (Citizen Science & AI Group: Chris Valentine, Mike Dodd, Janice Ansine, Stefan Rueger, Advaith Siddharthan, Damian Dadswell, OU UK)

This presentation shares best practice on the use, integration and testing of two artificial intelligence (AI) Cos4Cloud services in iSpot (www.iSpotnature.org).  Cos4Cloud is a European Horizon 2020 project boosting citizen science technological services to help increase and improve the quantity and quality of observations. It includes the participation of nine established citizen observatories, including iSpot, that have contributed to the development of these technological services. iSpot: your place to share nature is a citizen science platform (citizen observatory) for biodiversity developed and operated by The Open University (OU), a Cos4Cloud partner. It is aimed at helping anyone share wildlife observations, identify, explore and learn about nature.

With an interest in experimenting with automatic image identification to help the community identify observations and support learning, plans were put in place to introduce two AI systems on iSpot: Pl@ntNet-API (for plants) and FASTCAT-Cloud (initially for mammals then birds and invertebrates). Considerations for implementation also included whether adding an option that could more immediately give a response with a range of possibilities could be helpful as a learning tool for iSpot users.  Integration involved collaborating with the community to trial and test the services and user feedback was collated, to contribute to the development process. However, balancing user suggestions, demands, requirements and expectations could be challenging and this presentation will share incite from efforts to engage the iSpot community and user experience.

FASTCAT-Cloud is a service developed by DynAIkon that uploads and analyses nature videos and pictures, filtering out empty images automatically, helping to select more relevant images and recordings of wildlife activity from camera traps. The Pl@ntNet-API is an AI plant identification Application Programming Interface (API), developed by Inria, that uses Pl@ntNet’s image recognition to make identifications of plant species, which can be used to improve the user experience in citizen observatories.

Open XR Studios: What could immersive learning do for us? (Tyrrell Golding and Trevor Collins, OU UK)

Developing a sense of immersion has long been seen as an engaging approach to facilitate experiential learning. eXtended Reality (XR) offers opportunities to create immersive learning experiences using Augmented Reality, Mixed Reality, and Virtual Reality, which have been enabled by advances in the computer games and digital media industries. The Open XR Studios is an opportunity to embed the use of XR technologies within the module production and presentation process to create more immersive learning experiences for OU students.

With capital funding from the Office for Students, the Open XR Studios team are exploring how technologies such as 3D scanning, photogrammetry, 360-degree video, motion capture, volumetric video, and virtual production, can be used effectively within distance learning. This work builds on previous experiences from developing bespoke applications in modules to consider how XR can be deployed at scale as part of the OU’s core business. In this presentation we will introduce the project, our approach and initial use cases, and demonstrate some of the techniques we are currently developing.

Through this presentation we would like to engage delegates in considering the affordances of these technologies and how they could be used to enhance experiential learning within the OU. Much like the remote experiments in the Open STEM Labs enable students to undertake practical labwork at a distance, we want to explore how digital methods can extend reality so that distance learning students can access and experience places, procedures, and processes that enhance their learning.

The potential of generative AI in education (John Domingue and AI module team, OU UK)

The prevailing consensus among leading researchers is that Generative AI holds the potential to be as transformational as the printing press and the industrial revolution. In the realm of Higher Education, we have recently been conducting experiments with this technology to investigate its potential for enhancing module production and improving the overall student learning experience. Our endeavors thus far have encompassed the organization of four workshops aimed at understanding the present challenges associated with educational content production at the Open University and fostering brainstorming for innovative solutions. These efforts have yielded 149 identified challenges and 139 proposed ideas.

The ideas generated can be classified into six primary categories:

1. Accessibility – Automatically enhancing the accessibility or inclusivity of a module, such as through the automatic creation of image alt text.
2. Back-end – Facilitating the management of a module, for instance, by automatically generating metadata.
3. Learning experience – Enriching the learning process via means such as the automatic creation of virtual conversation partners or AI forums, enabling students to build confidence in their interactions.
4. Advisory – Providing support to staff or students through tools like virtual tutors.
5. Content creation – Automating the generation of unit conclusions or sample essays, among other resources.
6. Student journey – Automatically estimating study time required for each learner.

Additionally, we have developed a prototype platform where content creators can utilize GPT-4 to execute simple tasks associated with production, such as drafting an introduction or conclusion for a module unit. Our preliminary findings strongly suggest that the potential of Generative AI in Higher Education is immense, poised to bring about a radical transformation in the way we teach at the Open University.

Open Skills Academy (OU Wales) (Rhys Daniels, OU Wales)

Topics covered: AI, Large Language Models (GPT) & Education; Professional Learning; Widening access and participation in education with technology.

Background: The Open Skills Academy concept was developed by the OU in Wales under the work programme for the Research Wales Innovation Fund (RWIF) and strategy from the Higher Education Funding Council for Wales (HEFCW). The funding commenced in 2020 with one of the KPIs being ‘business skills and CPD’.

The problem: The way to achieve the RWIF KPI for business skills and CPD lay with promotion of OU short courses on OpenLearn, FutureLearn and the OU’s VLE. Capacity to do this with human resource is tight in Wales and there was no ability to recruit staff to promote the courses. Research into employer’s buying behaviour for CPD also highlighted key skills issues, including:

• 23% of employers experienced skills issues, but didn’t seek any advice (Skills Survey 2019)
• 9/10 employees in the UK estimated to need some form of reskilling by 2030 (CBI & McKinsey, 2020)
• 57% of decision makers say the skills shortage significantly affect their growth potential (OU Business Barometer 2021)
• 47% of employers want to do more, but can’t due to time, budget and inability to find relevant courses (Skills Survey 2019)
• 62% of employers in North Wales don’t engage with skills/CPD because “it’s too difficult to find what we need” (RSP Survey 2021).

A BDU project led also evidenced a desire from employers to be able to self-serve more.

The solution: To develop a ‘brand’ to be recognised as a one-stop shop for accessing short courses from each of the OU’s platforms.

This included the development of a search tool, powered by a knowledge graph and GPT, to enable people to enter search terms by skills, occupations and free text, to match to courses on all three platforms. Results are sorted by relevance and can be filtered by level, cost, and subject. The option to download results onto pdf also provides the opportunity for courses to emailed or saved for a later date.

Feedback: The search tool and user interface has been met with unanimous support from employers and stakeholders, delivering new income opportunities for the OU in Wales in the process. The average time to find courses requested from stakeholders and develop proposals is around five hours. Doing this through Open Skills Academy take around 20 minutes, delivering significant efficiency savings. The project has opportunities to grow commercial income, commercialisation, and further use AI to manage external relationships.

Ethical use of predictive learning analytics in distance education (Christothea Herodotou, Anna Gillespie and Irina Rets, OU UK)

The progressive move of higher education institutions (HEIs) towards blended and online environments, accelerated by COVID-19, and their access to a greater variety of student data has heightened the need for ethical learning analytics (LA). This need is particularly salient in light of a lack of comprehensive, evidence-based guidelines on ethics that address gaps voiced in LA ethics research. Studies on the topic are predominantly conceptual, representing mainly institutional rather than stakeholder views, with some areas of ethics remaining underexplored. In this talk we are going to present how we address this need by reflecting on four years of interdisciplinary research in developing the award-winning Early Alerts Indicators (EAI) dashboard. Through a lens focused on ethical considerations and informed by a practical approach to ethics, we conducted a case study review, using 10 relevant publications that report on the development and implementation of the tool. Our six practical recommendations on how to ethically engage with LA can inform an ethical development of LA that not only protects student privacy, but also ensures that LA tools are used in ways that effectively support student learning and development. Our talk will be based on the following recently published work: Rets, Irina; Herodotou, Christothea and Gillespie, Anna (2023). Six Practical Recommendations Enabling Ethical Use of Predictive Learning Analytics in Distance Education. Journal of Learning Analytics. DOI: https://doi.org/10.18608/jla.2023.7743