Trade unionists echoed their worries and those of those they represent, both for their jobs, future recruitment and the Welsh language. Some spoke optimistically about the jobs of the future and trade unions’ ability to respond to the challenge of AI.
Trade unionists echoed their worries and those of the individuals they represent. Through a series of focus groups and other meetings, they articulated anxieties related to their current employment, future recruitment practices, and the Welsh language. Amidst these concerns, some members conveyed optimism regarding the future job landscape and expressed confidence in trade unions' capability to effectively navigate the challenges posed by artificial intelligence (AI).
We have been told about increasing number of employers that are using artificial intelligence to come to a range of decisions about workers. That can even include choices regarding workers redundancies. As recounted by a trade unionist, his company employed artificial intelligence to determine the extent of job cuts, only to discover that the calculated figures were wrong. Stephen, an engineer and trade unionist, shared the following testimony with us:
“I worked in the electricity supply industry for 42 years. The company was bought out. I worked in the metering department. They decided to use an algorithm to forecast the volume of work and decide how many staff were needed.
They let us see the algorithm’s forecast. But, they didn’t show us how it worked. We did see that the figures were wrong and challenged them. This did help in mitigating the number of job losses.
I left the company at that time. But what I’m hearing is they realise they’ve made a massive mistake. They are now looking at recruiting people again. They were phoning people up, asking them to come back.
We were told by the managers ‘we’ve used the algorithm before, it’s tried and tested, it’s very accurate in its predictions.’
They just wouldn’t accept the logical arguments we presented to them about what the prediction was wrong.” (Stephen, electricity engineer)
The use of algorithms to decide on redundancy numbers even is extreme is not isolated: it is part of a trend towards the increased use of artificial intelligence in HR departments, replacing decision making mechanisms that involved line managers, and HR teams. We have been told of a broad spectrum of HR functions undergoing increasing automation, progressing through various stages of implementation, ranging from projects and pilots to smaller-scale instances and comprehensive roll-outs.
The various HR functions include some supposedly menial tasks such as summarizing grievances and disciplinary actions or drafting settlement agreements. The choice of this initial pilots seemed to the people we interviewed as a Trojan horses – seemingly uncontroversial, they open the space and lead the way to wider use of AI.
In other cases, we were reported the use of AI in creating contracts of employment, composing and analysing job descriptions, formulating interview questions based on job descriptions, and even generating job advertisements. But also in further areas of recruitment, including sifting through candidates or the projects to use software to monitoring speech patterns and microexpressions during interviews.
For example, during a meeting with union equality experts, George, a public sector rep told us that they had concerns about the use of AI to sift candidates. He said:
We've accepted for a while to ask questions that you can put a requirement on a job like ‘must have a degree or equivalent.’
So you can quite easily get a machine to do process whether someone has a degree from their completion of a form. AI can assess your application as to whether or not you have a degree. However, in my view, it gets more difficult when you rely on interpretation of subjective matters by machine learning, for example in assessing the answer to a question with much more nuance to it such as ‘how do you work with others?’”(George, public sector rep)
During the same discussion on the potential impact on Welsh speakers was considered. George works in a department which handled public consultations and monitored the public’s views on social media and said:
I have a lot of nervousness about how well Welsh language content is feeding into AI. From language models we've seen around the world they are good when you get content in the English language. But it is not so good at understanding Welsh particularly normal Welsh. That is, the Welsh that is used everyday as opposed to the formal textbook Welsh.
This could cause an issue when public bodies are using AI to monitor public sentiment on an issue. Whilst Bing Translate is quite good at formal Welsh it is not yet as so good at understanding tweets in Welsh where people will naturally be sloppier with their language. If we are not picking up that information that starts out as life in the Welsh language are we actually picking up what everybody in Wales is saying?”
This was echoed by others – they all agreed that automating the processing of language – not just public views as in this case - might lead to critical biases against Welsh speakers.
Whilst it is a new experience for workers dealing with public consultations, and other public service roles to be dealing with automation, it has been a fact of life for many years for those working in manufacturing. Reflecting on the increased use of computer programmes to control production lines and machinery, Eric, an experienced manufacturing worker and union rep from northeast Wales said:
If I had any kids, I’d advise them to study robotics. There are still going to be moving parts in factories, in fact there will be more. You’ll need guys to fix the automated machinery. They’ll need to be good at interrogating data. Engineering-wise and maintenance wise it’s going to get very interesting. (Eric, manufacturer worker)
His colleague Rory agreed saying
“In some ways it’s quite exciting. The youth of today are interested in technology and it could entice young people into our industry. The young are tech savvy.” On the other hand, he acknowledged that “guys my age are quaking, they’re living in fear.”
Rory’s comments point to the importance of providing high quality training to all workers in order to ensure that everyone can thrive in the modern job market.
The group also agreed on the importance of seeking to include clauses on new technology and AI in collective bargaining agreements.
It will be equally important that trade union reps are also trained on the new subject areas. David, a trade union tutor believed that the best way to ensure this was to build on existing course and materials.
"I think instead of trying to reinvent the wheel, instead of creating all brand-new stuff, let's have a think about what we are doing already doing. ‘When AI is the Boss’ and the e-learning tool, Managed by Artificial Intelligence are excellent.
We need to look at the activities we are currently running and is it possible to rewrite or to put something else into those activities? So, for example on the Unions Reps Stage 1 course, when we teach reps they should ask for information from HR about various different workplace policies - the disciplinary policy, for example, we could also put in a question to ask whether AI is being used within the workplace.
The same thing with health safety. When we ask reps to look for risk assessments, we should ask them was AI involved in the creation of these?”
In general, across the half dozen focus groups we convened, trade unionists emphasised their concerns about AI and its impact on the workplace. But it is important to record the optimism that some felt, and their determination to meet the challenges and opportunities head on.
Unions can take action now to secure benefits for workers in the new world of AI. Workers in the public sector can implement the WPC’s agreement on AI, automation and digitalisation in the public sector.
Meanwhile, workers in the private sector can seek to include clauses on AI in their collective bargaining agreements. Our report Negotiating the Future of Work: Automation and New Technology provides more information.
Between June and November 2023 we attended a series of focus groups and meetings with trade union members, reps and officers and reported on this in a series of blogs. In this blog we’ve reported on the views and experiences of trade union tutors, equalities officers and manufacturing reps. Views were shared on the basis of anonymity, so we have changed people’s names in this article.
Here are the other blogs in this series: