Diversity sourcing & AI: is it a match?

Hey there, it's Friday, and I've been thinking about how AI has already impacted talent sourcing since it's boom earlier this year.

But first, let's jump back to 2019.

I was part of a team working on a sourcing toolkit to help our european talent team boost up inclusive hiring practices.

The idea was pretty simple: let's create ready-to-use set of Boolean strings to spot underrepresented candidates based on their names and surnames for a specifically targeted part of the world. Sounds like a piece of cake, right?

But let me tell you, calling it tedious work would be an understatement.

We had to dive into statistics bureau reports and Wikipedia articles, scrape websites for the most common names in the region, and then think of some fancy excel fomulas to build up those Boolean strings.

After that, it was time for some manual testing. Even though we tried to add some automation into the mix, there always seemed to be room for human error.

And guess what? We stumbled upon a ton of issues - missing operators, typos, you name it. It was a total nightmare! The project, initially estimated to take three months, ended up taking around six months.

Now, in 2023, it's mind-boggling to think that our project could potentially be wrapped up overnight. Imagine all the time we'd save! Who knows where I'd be now if we had this project done quicker? ;)

What's the best take on such project these days?

1) In the current scenario, I'd probably start by reaching out to my friend, Chat GPT, Bard, or Bing chat (pick your favourite), and just ask them for some help.

Example prompt:

Can you build a boolean string from the top 100 female names in Poland born between 1950 and 2010. Please use an operator OR. Please make sure not to have any repetiotions.

2) Once I got the expected boolean strings, I'd give it a good look-over. I've already run into some made-up surnames and a few repetitions. It's a work in progress, for sure.

3) My typical approach involves asking ChatGPT for breaking down Boolean strings into bite-sized pieces to make it easier to read and use.

4) The last step includes testing booleans and potentially refining the propt to get expected results!

Yeah, it's not a rocket science, and I'm not another self-proclaimed AI expert.

The same idea of buiding diversity sourcing tools goes for creating Boolean strings based on specific criteria, for example:

• schools (e.g., girls-only or religious schools in the UK)
• organizations (e.g., female led industry-specific associations in Germany)
• feminine job titles in the languages you don't know
• languages and dialects used by communities in a specific regions

And the last, but not the least, I never put all my faith in AI.

As much as I love using ChatGPT to build my toolkits, I'd like to make sure that I keep my eye on potential bias affecting the outcomes.

Happy Boolean building, folks! ????

#diversitysourcing #diversityrecruiting #inclusivehiring