Upskilling Staff in AI

By Yusuf Berker, chemistry and computer science teacher - Sir John Deane's Sixth Form College

Introduction

Upskilling means expanding your existing skillset to enhance performance in your current role.

Will Al lead to such improvement for teachers? While that question remains to be answered, using Al effectively (or at least understanding its implications) is a challenge all colleges must take up. Here are some reasons why:

• Students will be using it, so staff will need to ensure they use it effectively and understand its limitations and relevant ethical issues
• Technology shapes society; proactive actions can help ensure Al has a positive effect on education
• Al misuse is classed as coursework malpractice
• It can potentially remove the repetitive aspects of creating reports and emails and analysing data

Opportunities

Personalised Learning and Feedback

In the UK, educational institutions are exploring Al's potential to tailor learning experiences to individual students' needs, preferences, and paces. For an example of how Al could be leveraged to personalise learning journeys it's worth taking a look at UCL's guidance to staff on Al use. While largely concerned with avoiding improper use of Al in assessment, it also explores the importance of understanding and integrating Al within education systems to better prepare students for the modern world.

Some contexts in which they suggest students can use Al in their own work include the following:

• drafting and structuring content;
• supporting the writing process in a limited manner;
• as a support tutor;
• supporting a particular process such as testing code;
• giving feedback or proofreading content.

These tasks often both require teacher involvement and are very time-consuming. If students could get some initial feedback and advice from an Al first, it could lower teacher workload. While there are potential issues, such as the Al giving unhelpful feedback or irrelevant information, the time saved may be worth it.

Efficient Admin

The application of Al in automating routine administrative tasks is widely acknowledged for its potential to free up educators and staff to focus more on student performance and other critical tasks. More time with students, and less on admin, can only be a good thing. However, teachers simply do not have the time to be researching tools in this area, and we would recommend that senior leadership and data teams carry out this research and identify, then thoroughly test, the most promising tools. Teachers did not invent PowerPoint, visualisers or spreadsheets and similarly cannot be expected to create custom Als for their own admin tasks!

Question Generation

Many an hour is spent on resource creation. This often involves creating multiple questions of a similar style and assessment criteria. For teachers who know what they want, and what good quality questions look like, Al promises to accelerate resource creation by instantly creating similar questions with new contexts. Mark schemes are more difficult to create, and tend to require significant editing. Here is an example process for using Al for resources:

1. Copy the original question into the Al chat
2. Specify exactly what you want it to do. For example, "write me a question in a different context but using exactly the same wording to that provided. It must be worth 4 marks and test the student's understanding of the difference between Turing Machines and Finite State Machines'
3. Feedback to the Al until you think the question provided is acceptable
4. Attempt the question yourself before sharing with students

This process may not initially be faster than just creating the questions yourself. But as you get better with prompting, you may find that it substantially increases your output.

Another advantage is that AI may create an unusual question (in a good way) that you may never have thought of. Since it contains knowledge of vast amounts of varied subjects, it could create unique scenarios that challenge students in new ways.

Risks

Skill Gaps

Without proper training, staff may feel overwhelmed by new technologies, potentially leading to resistance to Al tools. Some who are already unfamiliar with technologies such as spreadsheets may find it difficult to pick up. The training you provide must be clear, concise and applicable.

Staff should not be coerced into using Al, nor should it be a central part of teaching and learning CPD. Great teachers can be completely IT-illiterate! That said, it's worth pointing out to those who fall into this category that ChatGPT is easier to use than Excel!

New Teachers

There is an increasing plethora of freely available Al tools promising lesson resources made with a few clicks and prompts. While this may work for key stages that don't require highly precise content and explanation, at college level these tools come up short. New teachers should be trained by an expert in how to craft an effective resource or lesson plan; they simply will not learn this from current Als. In the question generation example above, a more experienced and expert teacher will be able to use Al more effectively - they will be able to see the flaws in initial outputs and have an idea of how to fine-tune them. As such, Al is no replacement for experience or training.

Ethical and Privacy Concerns

All data provided to an Al is stored somewhere, at least temporarily. It's important that staff 'clean' any data before uploading it anywhere, especially student data including names and qualifications.

Case Studies and Best Practices

Question-generation

Staff in my sixth-form college have been experimenting with using Al to create new questions, as discussed above, including multiple-choice questions. If successful, this can rapidly speed up question generation and also provide unexpected contexts that teachers may not naturally think of. This is most useful in subjects that are relatively new, or very niche, where the number of quality questions available is low.

Subject-specific Presentations to Students

It is vital that each subject tells students what is and isn't acceptable Al use in their contexts. This is especially important in coursework subjects. Departments should meet and discuss the JCQ guidelines to carefully prepare their own advice. Any Al use that isn't cited in their work Is currently classed as malpractice so students should be encouraged to save all their interactions.

Streamlining Administrative Tasks with Al Chatbots

The University of Murcia in Spain saw significant success after implementing an Al-powered chatbot to handle student inquiries. This chatbot was able to answer questions with over 91% accuracy, handling 38,708 questions effectively. This implementation provided round-the-clock assistance to students without any changes to existing staff structures.

ChatGPT (paid version) now provides the option for you to upload your own documents to a bot, provide some instructions, and have it answer questions based on the resources you gave it. This could potentially make teachers' lives easier, especially over the holidays, as students can get standard answers to simple questions from the bot.

Intelligent Tutoring Systems

Many students seek additional help for subjects they find difficult. An increasing range of Al tutors are becoming available, perhaps the most notable being Khan Academy's version KhanMigo. Their Al is trained to not provide students with answers, and will let the student lead the discussion by asking them questions. If the student is trained to use it properly, they could get a lot out of a tutor with a bottomless amount of patience and analogies up its sleeve. Currently only the Lite version is available in the UK, but it is still worth a try.

Key Messages for Leadership

It is tempting to rush into using Al just because everyone else Is. Don't! Here are two key questions to ask yourselves before embarking on any AI CPD:

• "What problems are we looking to solve?"
• "What are we trying to achieve?"

If you do decide to embark on Al training, here are some suggestions. All training should be practical and you must explain the why behind it.

Al Workshops: The best place to start would be an introduction to the fundamentals of how the language models work and some suggestions for using them in question generation. Get teachers in computer rooms doing it for themselves!

These sessions can cover a range of topics. The most important are likely ethics and data security. An eager staff member trying to use ChatGPT to do data analysis could upload a spreadsheet full of confidential information without thinking.

Pilot Projects: Ask a willing staff member or two to introduce Al tools into their teaching or admin and invite them to share their findings in a presentation to staff. It's much more likely to gain traction and be useful if staff in the college can share real examples, including areas that they found to be unproductive!

Key Messages for Students

Al is a Partner in Learning

Emphasize to students that Al tools are here to enhance their learning experience, offering personalized support and feedback. It must not do the work for them.

Here are some potential uses students may find it helpful for:

• Ask it questions about topics you have struggled with. You can request analogies and examples to make the content more accessible to you.
• Use Al in the early stages of a project to suggest some ideas and help you structure your work.
• Get some general feedback on written work, including clarity of communication and structuring. It won't be able to provide the subject-specific feedback your teachers can give you, but it's a good start.

Al is not a perfect haven of knowledge.

Encourage students to critically assess the role and impact of Al in society, preparing them for a future where Al is used in all areas of work and society. Here are some issues they could consider:

• Bias from the data the Al are trained on
• Lack of referencing in their outputs
• What skills will become more important for careers as Al becomes more prevalent
• Impact on employment of automation

Conclusion

At the time of writing, there is no silver-bullet software that takes Al and applies it to teaching in a way that is accessible for teachers of all levels of technical ability. Teachers should consider whether they have any problems that Al could solve before jumping in to use it. The most promising test cases are developing questions where there are limited resources available, and/or where the teacher is part of a small department. The next area to explore is data analysis, with ChatGPT now allowing spreadsheets to be uploaded and analysed. In all areas of development, data protection and ethics must be a priority, with staff trained on how to use Al without compromising confidentiality. Once teachers and other staff are routinely, effectively, and ethically using Al in question generation and data analysis, they are likely to be equipped with the skills to identify further possible uses. Building a strong foundation by doing is key to staff applying themselves to Al with greater autonomy.

Reading List and Resources:

• aiforeducation.io - Website offering a foundational understanding of Al in education and very helpful pictographics.
• Education Technology Association - A website with resources, case studies, and best practices for integrating Al in education.
• Future Proof: 9 Rules for Humans in the Age of Automation by Kevin Roose - A book that explores how we can develop skills to make us more competitive in the age of Al.
• Human Rights, Robot Wrongs by Susie Alegre - A book that critically approaches the development of Al and its impact on human rights.
• Artificial Unintelligence by Meredith Broussard - argues that enthusiasm for technological solutions can lead to inappropriate uses, and that not all problems require them.
• 'Reskilling in the Age of Al' in Harvard Business Review magazine (September 2023): What BCG and Harvard researchers learnt from talking to 40 business leaders about preparing their staff for the Al age.