This AI will tell you whether you're being paid enough (or too much)

How helpful would it have been for you to know the salary of that job, or at least the potential salary range, before you went through the whole application and interview process? There's an AI for that.

It's evident that employers are reluctant to reveal salaries at the application stage, but researchers at the Stanford Institute for Human-Centered Artificial Intelligence are helping circumvent this barrier with the help of AI, so prospective employees can decide for themselves which job listings are worth their time (via Venture Beat).

The process involves training the AI on a dataset of over a million job postings, which allows it to "use the text of job listings to evaluate the salary-relevant characteristics of jobs in close-to real time." 

It does so with 87% accuracy, compared to the 69% it got from using the job postings’ titles and location alone.

A study of 6,000 professionals in the UK showed 81% of us cite salary as the most important factor in finding a new job, and yet the earning potential transparency issue persists. Employers simply don't want to give you a heads up on compensation, despite the fact that job postings with salaries listed tend to yield twice the number of successful candidates.

The BBC reports that, back in 2019, only 12.6% of companies worldwide revealed the salary range in job adverts. 

That number is growing, but not nearly fast enough. The report goes on to explain that companies that are "forthcoming about their wages can attract better, more diverse talent, making salary transparency an actionable way of creating a more equitable workplace."

This year on International Women's day, the UK Government announced an initiative to improve transparency in the application process, to help women negotiate fairer salaries. So, steps are being taken in the right direction at least. But what if there was a way to predict the salary of a job, just by analysing the wording of the job posting?

Sarah Bana of the Stanford Digital Economy Lab is the one training AI to help predict salaries like this. Her research (PDF warning) proves that since "text in job postings is written in commonplace language," its entirely possible to predict their salary using a technique called "transfer learning."

Essentially, the AI could give prospective employees more chances to pick and choose jobs, rather than applying for everything, only to be disappointed. She says it'll "make applying for jobs more transparent and improve our approach to workforce education and training."

Sounds grand to me, and it'll even give people more insight into the value of upskilling, by detailing which skills are the most highly sought after by employers. So, anyone deciding on a future career path can hone their skills according to how much dollar they can earn from them. 

Ah, the marriage of AI and capitalism at its finest.