Working for the Man
  • 4th July, 2019
  •  

Working for the Man

By Amy Bray

3 min read

Personalisation is the holy grail of modern customer facing businesses, like banks, retailers and streaming music services. However the extent of personalisation at most organisations is no better than your name inserted into an email. Online news feeds however, are several steps ahead. I’m sure most of you would have noticed how your news feeds are tailored to you and eerily to your most recent searches. But how is this digital sleight of hand actually done? It turns out to be a combination of clever algorithms and - you.

Google’s news feed for example, uses a balance of machine learning (ML) and user relationships to personalise your news feed. The machine learning part of this looks at your Google Activities*, which may include:

  • Web activity - what you have Googled
  • Location history - Google Maps
  • Device information - your calendar and contacts
  • Voice and audio activity - from using “Hey Google…”
  • YouTube search and watch history. 

But using machine learning alone isn’t sufficient because the algorithms will sometimes get it wrong. Just because you clicked on a link doesn’t mean you are necessarily interested. You may be drawn to an obscure aspect of a video (e.g. the camera angle), rather than the topic of the video itself (e.g. Italian cooking). Machines struggle to pickup these subtleties and this is where manual personalisation comes in.  

If a story on News24 was about the State of the Nation, Google allows you to choose “not interested” in either “State of the Nation” or “News24,” giving you a way to control your personal feed. You might also be encouraged to reveal your preferences through the “Did you find this story useful? Yes/No” question that occasionally pops up. This new information is in turn used to “teach” the underlying algorithms about your personal preferences.

Another smart or insidious tool, depending on your point of view, is Aggregation where you might be shown multiple articles about the same event. If you happen to be following your favourite team’s performance at the Cricket World Cup, you are likely to see several articles about cricket that are grouped together. Articles from different editors and authors often give different slants on the same topic, allowing you to home in on the one you are most interested in - and once again providing information on what content appeals to you the most.

The majority of us don’t know what information the internet holds on others, let alone ourselves. Yet we inadvertently add to this treasure trove of data every day. 

References:

https://dzone.com/articles/google-feed-personalization-and-recommender-system

Share This Artcle :

About Author

Amy is a physics graduate. She is now entering the world of data science with Ixio and Zindi. She is originally from Durban but is now making a life for herself in Cape Town. Amy loves a good problem and enjoys being part of the whole solution process.

Comments