Generative Humans and AI in Early Careers

Breakfast News on 28 September 2023, was the 3rd AI event that GTI has run over the last 3 years – well before it was elevated in our collective consciousness with the launch of ChatGPT in November 2022.  It understandably remains a popular topic in our sector with 150 people registering for the session in London and 800 online.

This is a summary of the content from the event – bringing together the speakers’ content into themes and, in particular, referencing their frameworks to help make sense of the topic. You can access the slides, speaker and student panel videos from the event below.

‘Automated rules-based AI’ vs ‘Generative AI’

AI is nothing new. It refers to ‘automated rules-based decisions’ - like a Chatbot or credit score - which have been around for decades. So, what’s changed? It’s the sophistication in Generative AI…the idea of a discussion chat-based interface where you can ask anything using natural conversational phrases or upload images or documents. Information, data, video, audio or images are generated in near real time that resembles ‘human-created’ content. This means that we don’t need to be ‘good at Googling’ or be a programmer to get sophisticated results.

So Generative AI tools can produce new content tailored to the needs of the user or organisation and do so in an iterative and interactive way. Tools like ChatGPT, Claude, Google Bard and Dall-E are leading the way.

If a software provider platform claims to use ‘AI’ (which many now do), it is worth clarifying whether this means it uses automated rules-based decisions (much software can be thought of as using AI in this sense) or Generative AI which is newer in the way it delivers for users.

Why ‘Generative Humans’?

If AI is increasingly automating routine tasks to support skills development and aspects of the hiring process, all of us in the early careers sector need to develop our skills to become ‘better humans’ in areas that aren’t automatable. In areas like the design and delivery of skills workshops, careers fairs, insight events, assessment centres, candidate experience design (CX), candidate care, coaching and role modelling. There will also be a need for expert human oversight of automated decisions.

A fast start for students but slower adoption by educators and employers

The student panel at Breakfast News commented on their use of Generative AI in areas like:

• Revision scheduling. Helping with planning ahead and setting deadlines.
• Outline of essays.
• Facilitating learning. Asking quick questions like ‘what is the difference between term A and B in layman’s terms’.
• Research. Asking about a company to consider whether to apply or to help in the selection process.
• CV improvements. Refining the bullet points and tailoring to the employer or industry sector.
• Helping (or cheating) in exams and assessments. This is discussed in a section below.

Looking at employer usage, Stephen Isherwood, Co-CEO of the Institute of Student Employers (ISE) highlighted that on recently asking 92 large employers in the UK whether they are currently using AI (meaning automated decisions not Generative AI) the answer is that only 9% are. Finance and Professional Services firms are the most likely to be, followed by Digital and Tech.

This contrasts with the view held by many students, and sometimes our sector, that most large recruiters make automated decisions. If only around 9% of recruiters are making some level of automated decisions, it is likely that a lower percentage are using Generative AI – showing the potential for much more use over time.

ISE has found that the most common uses, in order of usage are to:

• pre-screen candidates
• analyse video interviews
• screen CVs

AI and Generative AI applications in use, or likely to soon be used, include:

• Skills understanding. Universities and employers can generate role and skill taxonomies. Tristan Mathieson, from GTI’s Data Team, showed how GTI is working with Leeds Arts University to analyse the creative and digital skills that employers are looking for.
• Chatbots. For educator curated careers advice and guidance. And for employers to answer questions through the application and onboarding process.
• Careers guidance. Providing tailored learning content or re-writing existing content in a more accessible format. Testing learning progress. Advising on career paths and offering coaching.
• CV reviews. Although most large employers have removed CVs from their processes, the majority of SMEs use them and the majority of graduates work for SMEs. Students can use Generative AI to create different versions of their CV, highlighting the skills and achievements most relevant to the role. Careers teams have a role to advise, and automatically (using a CV review tool), or manually, review CVs and suggest improvements.
• Job descriptions. Employers can create job descriptions tailored to the needs of their organisation, considering current skills or skills gaps.
• Candidate Attraction. Employers creating more personalised emails to engage students.
• Personalised assessments Creating candidate assessments and interview questions based not only on the job requirements, but also on a candidate’s profile and experience.
• Hiring decision support Pre-screening candidates, reviewing video interviews and screening CVs. It should be noted these are some of the ‘original’ uses of automated decision making and remain some of the highest risk in terms of bias.
• Transcribing interviews. Automatically writing up the discussions.
• Candidate feedback. Traditionally via automated assessment reports, feedback can be further personalised and presented as more a supportive (chat-based) discussion.
• Onboarding. Personalised onboarding experiences, like answering commonly asked questions, providing training materials, setting up meetings with stakeholders, and helping to navigate complex HR issues.
• Training careers and employer teams. Tristan Mathieson, from GTI’s Data Team showed a simple internal Q&A tool developed to answer internal questions through a natural chat interface.
• Team engagement. For example, drafting survey questions, analysing sentiment, summarising findings, recommending actions.
• Learning and development. Similar to Careers Guidance (above). Including drafting performance objectives and reviewing progress.
• Pipeline reporting and learner analytics. A chat interface to your data. Tristan Mathieson, showed GTI’s ATS reporting (below).

Thanks to a Venero Capital Advisors 2023 Report for some of these use cases.

Future of Work, AI and Data Science Skills

The fear that AI will eliminate jobs is real – and nothing new. As technology advances, new career options are created. Students are generally aware of the potential of automation and the possibility of job displacement. They also recognise the emergence of new and potentially improved employment opportunities and the need for AI related skills development.

A recent JISC report here highlights that some students are concerned that they are the cohort that is missing out on developing AI skills that they may need for employment. Students are seeking support from educational institutions to bridge the gap.

Understanding the importance of adaptability, one of the Breakfast News student panel explained that in his aerospace engineering degree, the emergence of ChatGPT has encouraged him to choose more technical modules to help future proof himself. But overall, he is excited about AI’s potential.

We’re seeing employers more highly prioritise human skills and a candidate’s ability to work with the help of Generative AI. GTI and Cappfinity have developed the https://skillsforskills.org/ framework to define these skills and the idea that the future of work will be about always learning and adapting.

When considering technical Data Science and AI skills, employers are starting to want higher levels of skills, as AI tools and APIs to them mean that less skills are needed to do the ‘basics’. This has implications for short courses and bootcamps.

GTI is supporting the Department for Science, Industry and Technology (DSIT) and the Office for AI (OAI), working with the Office for Students (OfS), to support diverse students to study AI and Data Science at postgraduate level. This will boost UK innovation and help to reduce the risk of bias in algorithms.

Employers are invited to sponsor one or more students that may otherwise not be able to study at this level. Viscount Camrose, the Parliamentary Under Secretary of State (Minister for AI and Intellectual Property) discussed the scheme (below). To set up a discussion, please leave your details here.

AI in Tests and Assessment

Integrity can be defined as ‘demonstrating a moral compass’. In light of the advances in Generative AI, will integrity in exams, online assessments, and assessment centres now be part of the cultural norm? In other words, should using tools to pass tests be defined as cheating, or not? The answer to this question may evolve over time.

Cappfinity provides the following framework when thinking about cheating in tests. Nicky Garcea, Co-CEO's talk is below.

Deter

Employers often don’t, but can easily, explain to candidates in a well-crafted sentence(s) that they shouldn’t cheat. As one of the student panel said, he avoids cheating as he wouldn’t want to take a job if he wasn’t confident that he had the right skills. Another panelist said that he would appreciate guidance as to whether to use AI and other tools, on an assessment-by-assessment basis.

Educators have offered guidance on plagiarism for some time, so there are lessons there. The majority of students on the panel said they would appreciate more guidance from their university. One student explained that their university has said detailed guidance is a work-in-progress so avoid the use of AI tools for now.

A second approach to deterring cheating is offering ways for students to prepare by practising. When we’re confident in ourselves and feel we’re good enough, we're less likely to cheat.