Making meaningful algorithmic recommendations require that your algorithm is “fed” with data about your customers tastes.
This can be done in many ways :
- implicit feedback : it’s when you track a user’s actions and infere her taste based on them
- collaborative filtering : using someone’s profile and other similar persons’ taste you are then able to make recommendations
- explicit feedback : when you let users the possibility to tell you in an explicit way what they want, what they like and dislike (for further information on the importance of explicit feedback read this earlier article)
I’ve found recently two interesting examples of explicit feedback mechanisms
2-step implicit feedback on CNN mobile website
Let’s start with the CNN website. Although a “Read More” button has been in use for quite a long time on many different websites to track readers genuine interest in a piece of content, CNN “Read More” button is slightly different and I don’t know why.
When hit once the article won’t show. rather the button will turn black and you’ll have to hit it once more to read the entire article. Is it a failure-proof version of the 1-click version?
If someone has an idea I’m happy to learn.
Snooze Trump function on the Quartz mobile app
This one is a must see.
Let me first say that I love the Quartz mobile app. It’s so different from all other news apps you can use with its chatbot look and its way to interact with you.
The Qz.com app has an amazing explicit feedback mechanism implemented in the app setup. You can snooze all news about Donald Trump during 24h. Isn’t this a strong signal ?
Tags: algorithmic governance, market research, recommendation algorithms