gilescolborne: How I stopped worrying and learned to love algorithms. https://t.co/iYFZMDFD70 #uxlx


Spotify has a feature where they suggest a playlist to the users, based on their algorythm. What did the designer do to add value?

Natural Languages interfaces are being used more and more.

The GUI doesn’t need to be redesigned.

Will designers be replaced by data scientists ?

How algorithms work

  • ranking
  • sensory input
  • anticipation and agents
  • natural language

They need data to work.

First layer of data: time table data for example. Second layer: gps data from the bus Third layer: local weather, traffic…

we can use the different layers to anticipate and predict what will happen.

Inputs —> algorythm —> output

Has a designer you need to know what output is useful for the user.

You start with a raw algorithm and it evolves being adjusted as time goes by.

gabrielhl: Watching @gilescolborne talk at @uxlx? Nice long reads on Spotify Discover Weekly: https://t.co/3yLGVkTIzV and https://t.co/502Unvh5a3 #uxlx


Don’t just add extra layers, find which make the most difference to the user to find the optimal solution faster.

There as to be a dialogue between the designer and the engineer to find what is best to the user. — Giles Colborne

Conversations can’t start with a blank screen and need to setup a path to success.

Just because the answer is correct, it doesn’t mean it’s the right answer. What format does the user want to see?

stephanierieger: Basic algorithm design: Inputs (what’s available, accurate, meaningful) + Algorithm (w/ relevant training) + Output (what user needs) #uxlx


users need a clear signal: are they talking to a bot or a human ?

Spotify’s discover weekly feels like a mixtape from a friend and not a message from a bot.

There is an undefinable quality between those tracks.