translate this page >
volunteer

Artificial Intelligence (AI) and Recruitment

Artificial Intelligence (AI) and Recruitment

AI and Recruitment

Artificial Intelligence (AI) and recruitment. “A match made in heaven!”, some will say, “A match made in hell!”, others will retort. Volunteer Edinburgh’s Employability Service has worked with clients who have successfully used AI to assist with their job search and with clients who had to sign that they didn’t use AI in the application process. So, should you or shouldn’t you use it? We think it’s worth taking a closer look to understand the pros and cons and to make an informed decision.

When we talk about AI these days, one of the first things mentioned is ChatGPT. Released in late 2022, it caused a sensation with many users being equally concerned as fascinated and occasionally amused. Ask the application to write a shopping list in the style of Robert Burns and it’ll comply to its best abilities. We leave it to the reader to judge the quality.

Upon yon parchment fair and plain, A list of goods, a humble gain. From field to store, my steps do tread, in quest of items, here are spread. […]

For those who haven’t had the pleasure yet – ChatGPT is an AI chatbot. You might have encountered chatbots on websites in the form of little chat windows that pop up – “Hello, how may I help…?” The difference is that these “normal” chatbots are trained to provide instant answers to queries with specific, limited datasets. Asking them to write a Burns inspired shopping list would fail.

ChatGPT on the other hand is based on a so-called Large Language Model (LLM). It is trained using enormous datasets. This enables it to operate in a way that is much more advanced and sophisticated. It possesses the ability to create and analyse and, to a certain degree, self-improve.

ChatGPT is also equipped with advanced Natural Language Processing (NLP). Imitating human communication, users basically type their requests as they’d speak. This makes the interaction feel more natural. Now Natural Language Processing might sound a bit exotic, but in reality it is familiar to most of us. Think about your mobile phone, Siri and Alexa, these are all systems that are using NLP to action tasks based on voice commands.

Most of Volunteer Edinburgh’s Employability Service users state that searching for a new job is a job in itself and can, at times, be overwhelming. Using available tools and services to ease the process is most certainly tempting. This includes employability services, CVs and cover letter templates, grammar and autocorrect for sure but should or shouldn’t it extend to using AI as well? Although technically a form of AI, most people would probably agree that using autocorrect and applications like Grammarly is still ok. So why not upload a CV and job description and ask ChatGPT to write a cover letter? “That’s cheating!”, many will say. “It’s just a starting point!”, others will retort.

But isn’t it just fair play given that an increasing number of employers are using AI-based solutions at their end?

Large and mid-size companies dealing with high numbers of applications are known to have moved towards tech/AI-based hiring solutions. Today, applicants need to be aware and prepared that the processing of their application may be partly or completely AI-led. Applications may be reviewed by AI screening tools, the shortlisting for interviews may not involve any human interaction/revision either and even at interview stage candidates could face an AI-led Automated Video Interview.

Whilst AI is most certainly allowing employers to streamline processes and save resources, some also believe that AI is less biased. Others raise concern that it isn’t transparent what data these applications are basing their decisions on and that the data they use could result in bias. LLMs are trained on already existing datasets. The biases contained in these datasets could potentially be continued and replicated if not corrected. For example, if historically certain demographics were under represented in a company, the AI might continue to select candidates from those demographics and exclude others.

Furthermore, a Harvard Business School report highlights that AI systems are trained to limit risk and select conservatively if applications don’t fit the mould. As a result candidates who might be suitable for a different role than the one they applied for will simply disappear leaving high talent untapped.

But what would be the risks of using applications like ChatGPT as an applicant? Firstly, it is a matter of conscience. Some companies want to make sure that the application is a true representation of a candidate’s skills and will ask you to confirm that you haven’t used AI in the process. Beyond that, cover letters generated by AI can be generic. They have a lot of big words but lack personality as well as reference to specific examples. AI could potentially be used as a starting point but it would be recommended to personalise the result and fact-check it. It is known that AI executes requests but doesn’t necessarily question prompts. It would happily write an article on how Scotland is the country with the least amount of rain on earth as you can see below.

Scotland’s newfound status as the world’s least rainy country challenges preconceived notions about its climate.(…) As the world becomes more aware of Scotland’s dry side, the country may witness a boost in tourism and a reevaluation of its agricultural potential, ultimately reshaping its image on the global stage.

Another point is that employers might actually check whether a motivational statement or a cover letter was created by AI. This would most certainly not be the first impression any applicant could want and potentially mean disqualification. On the other hand, it is actually still quite tricky to prove whether a piece has been produced with the help of AI or not. There are a few services but their accuracy heavily depends on text lengths and apparently on whether or not the author is writing in their first language.

Stanford assistant professor James Zou and his team published a study that indicated AI detectors having a dramatically low accuracy for texts written by non-native speakers, falsely misclassifying them as AI generated. Therefore, AI detectors results can’t (yet) be dependably relied upon. Lastly, applications issue warnings themselves that at times the datasets they have been trained with might not be up-to-date and that users shouldn’t share sensitive data whilst using applications.

In the end it comes down to personal preferences. Some of Volunteer Edinburgh’s clients reported their appreciation of AI input/feedback on grammar and structure of cover letters as a starting point. Others stated that they waived AI support to comply with application requirements and enter a job interview with the knowledge and confidence that their knowledge, presentation and communication was consistent with their application.

Whether you are a fan or a casual observer of AI, it is clearly playing an increasing role in how employers recruit staff – and in how job seekers find work.