June 14, 2018

Bias in algorithms



Is your computer prejudiced? It might sound strange, but the answer could be yes.

Algorithms have lead to some incredible advances for humanity. But they’re ultimately based on the data humans feed into them, and that can include some pretty ugly stuff. If digital technology is a mirror held up to society, it reflects our society’s dark side too.

If you’re looking to improve your business or non-for-profit’s SEO strategies or social media reach, it’s useful to be aware of how algorithms can be biased – and why it matters.

The problem with algorithmic bias

These are the results from a search for “professional haircut” on Google:

results from a search for “professional haircut” on Google

…and these are results for “unprofessional haircut”:

results for “unprofessional haircut”

Notice a difference? Now, Google isn’t deliberately coded to be racist or sexist. But it is coded to uncritically reflect the beliefs and expectations of a society where racism and sexism are still systemic problems.

And even when an image or search result isn’t problematic in its original context (beyond the images of this second example is actually a collection of think pieces of why these should not be considered examples of unprofessional haircuts), Google doesn’t always pick up that subtlety.

This has massive consequences, because search engines shape how we see the world in all kinds of subtle and not-so-subtle ways. They can reinforce prejudices and build new ones, and the fact that we tend to think of them as objective only makes them more dangerous.

If you Google “Sudanese Melbourne”, one of the top images is a photo of minors being pepper sprayed. If you click through, you’ll find yourself on a white nationalist website.

Search results have also been proven to impact the decisions of undecided voters by about 20%. But that’s not the only area where algorithmic bias can do damage.

Historically, Facebook algorithms have done little to differentiate between credible journalism and clickbait or “fake news.” Facebook ads can be targeted to cut out particular social groups, with sinister implications. Experts have claimed that Linkedin’s search results may show gender bias. And who could forget Microsoft’s short-lived machine learning Twitter bot Tay, which was tweeting out hate speech within hours of going live?


We’ve spoken about the extreme end of the scale, but bias algorithmic bias something everyone who’s interested in improving their SEO needs to think about. Steph blogged about what to do when your keywords run into legal problems – but what do you do when they’re unethical or bad taste?

Often, the correct or sensitive term for something isn’t what people search for. For example, we recently worked with a provider of residential aged care facilities. Except people don’t search for “residential aged care facilities” as often as ­they search for “nursing homes” – a term the industry steers clear of because it sounds demeaning.

In the end, the client chose not to use the SEO-friendly phrase “nursing homes”, because respect for their customers came first.

It’s not worth compromising your company’s values for a sneaky SEO win. SEO is important, but getting clients or customers to your page isn’t enough ­– you have to give them a reason to stay there, and prove that your organisation has integrity.

That’s why high-quality content is so important. At Avion, we take pride in writing search engine optimised copy that supports and connects to readers. If you’re interested, get in touch with our team of Melbourne copywriters – we’d love to have a chat.

Industry Insights

January 21, 2022

Content design in theory and practice – who’s doing it well?

November 26, 2021

Who writes your copy? Spotlight on… Marina Penderis

November 10, 2021

Fintech 2021: Who survived and what’s ahead?

October 29, 2021

Could you use a microsite for your next campaign?

October 6, 2021

5 ways technology’s impact is a force for good