Let me remove things your AI shouldn't use for recommendations
Daniele Catalanotto
May 18, 2022
A Service Design Principle inspired by Guy Martin
Watch me think out loud about this new principle
The original idea of Guy
My frustration with this regarding Spotify, is that they don’t allow you to remove previously played songs from the recommendation algorithm. (Maybe this is another principle?)
Amazon builds recommendations based on things I’ve looked at or purchased, but I can look at this history and remove items at will so they won’t be used to make recommendations. Knowing this exists means I feel more confident looking for items that I need but might be embarrassing, as I can just excise it from my history afterwards.
If I can remove the 37 plays of “Space Unicorn” from my Spotify history, I would be a happy man.
Amazon builds recommendations based on things I’ve looked at or purchased, but I can look at this history and remove items at will so they won’t be used to make recommendations. Knowing this exists means I feel more confident looking for items that I need but might be embarrassing, as I can just excise it from my history afterwards.
If I can remove the 37 plays of “Space Unicorn” from my Spotify history, I would be a happy man.
Side notes
- This is the first draft of this Service Design Principle.
- Once adapted and refined multiple time, this principle might be part of the book "Service Design Principles 201-300".
- Let me know if you find this thinking out loud format also interesting for you.
8 comments
Guy Martin
May 18, 2022
Thanks for the shout-out Daniele!
Space Unicorn is up to 85 plays now... Spotify are you watching?
Space Unicorn is up to 85 plays now... Spotify are you watching?
Daniele Catalanotto
Jun 4, 2022
Hahaha Spotify's algorithm is definitely thinking: what a special taste in music this guy has :P
Daniele Catalanotto
Aug 10, 2022
So here comes a second version of this Service Design Principle but this time in written form:
You are browsing on Amazon with your partner to buy something. Suddenly your partner sees that Amazon recommends you some jewellery based on your last purchase. She smiles. She now knows that you have bought her some lovely jewellery for your couple anniversary. But, you, you are sad. Your surprise gift isn’t any more a surprise.
Guy Martin (1) explains that Amazon has a smart feature to avoid precisely those types of moments. In his Amazon profile, he can go into the history and then say that remove something from there. Now, this secret gift will not be used for recommendations in the future.
This smart move is not only great for keeping surprises lovely. It can also come in handy for our guilty pleasures. Guy says it well:
“Knowing this exists means I feel more confident looking for items that I need but might be embarrassing, as I can just excise it from my history afterwards.”
You maybe have a guilty pleasure that you don’t want to get reminded about in the future as you are changing your lifestyle. Like someone who removes all cookies from home to not get tempted by sweets while trying to lose weight. Nobody would like to have someone knocking on the door and saying: “Hey, I think you’d like some chocolate as you did in the past!”
It’s also smart when buying or watching something for someone else. For example, I don’t want to get recommendations for word puzzles, even if I’m the one buying them for my wife.
What’s true for a shopping experience can also serve in other sectors. For example, Guy explains this about his favourite music streaming service: “If I can remove the 37 plays of “Space Unicorn” from my history, I would be a happy man.”
As more and more services create recommendation systems, users should be able to say what type of stuff they don’t want to get recommended. In practice, I see three different ways to give people back their power:
Like Spotify or Netflix, you can ask people to rate what they are listening to or watching to recommend more of this stuff.
Like Youtube, you can let people improve your recommendations with a feedback button. With it, users can tell you if a recommendation is good or bad for them.
Like Amazon does for the pro users, you can really go deep and let people remove things from their history to fine-tune your recommendations.
So let me ask you:
What would be some smart and valuable recommendations you could give automatically to your users or staff members? How will you let people customize these automatic recommendations?
Footnotes
(1) A smart shopper and a member of the co-creator Service Design community who inspired this story. Thanks Guy!
Daniele Catalanotto
Aug 10, 2022
I made a video and a short note about the behind the scenes of the writing that lead to this second draft.
Daniele Catalanotto
Aug 13, 2022
The third draft of this Service Design Principle
You are on a shopping site with your partner to buy something. Your partner notices that the site recommends you buy some jewellery. She smiles. She now knows that you bought her some lovely jewellery for your couple anniversary. But, you, you are sad! Your surprise gift isn’t any more a surprise!
Guy Martin (1) explains that Amazon has a smart feature to avoid precisely those types of moments. In his Amazon profile, he can go into the history and remove something from there. Now, this secret gift can’t be used for recommendations in the future.
This smart feature can also come in handy for our guilty pleasures. Guy says it well:
“With this feature, I’m more confident looking for items that I need but might be embarrassing. I can just excise it from my history afterwards.”
You might have a guilty pleasure that you don’t want to get reminded about. Like someone who removes all cookies from home to not get tempted by sweets while trying to lose weight. Nobody would like to have someone knocking on the door to remind you of your past bad habits.
It’s also smart when buying or watching something for someone else. For example, I don’t want to get recommendations for word puzzles, even if I’m the one buying them for my wife.
What’s true for a shopping experience can also serve in other sectors. For example, Guy explains this about his favourite music streaming service.
“If I can remove the 37 plays of “Space Unicorn” from my history, I would be a happy man.”
More and more services create recommendation systems. Maybe users should be able (2) to say what type of stuff they don’t want to get recommended.
So let me ask you:
What are smart and valuable recommendations you can give automatically to people? How can you let people customize these recommendations?
Footnotes
(1) A smart shopper and a member of the co-creator Service Design community
(2) Here are a few other ways to give people back their power that I’ve seen.
On Spotify, you can rate what they are listening to or watching to recommend more of this stuff.
On Youtube, you can improve your recommendations with a feedback button. With it, users can tell you if a recommendation is good or bad for them.
On Netflix, you can create different profiles. And maybe create one only for your guilty pleasures. To explore why profiles are a smart idea, read the principle “Remember that I have multiple roles in my life.”
Daniele's personal notes
- This is the third draft of this Service Design Principle
- I've been able to reduce by 28.66% the length of this draft compared to the second version.
- I've put the other examples from Spotify, Youtube and Netlfix in the footnotes. Maybe in next drafts, I'll completely remove these mentions.
- I've added a footnote that mentions a story that is pretty similar but that is more about the different roles we play in our lives.
- I wrote a detailed note about how I use "Hemingway Bridges" to make the book editing easier for me while using this principle as an example.
Daniele Catalanotto
Oct 31, 2022
The third draft of this Service Design Principle
You're looking for new furniture online with your partner. Your partner notices that the site recommends you buy some jewellery. She smiles. She now knows that you didn't forget your couple's anniversary.
But, you, you are sad! Your surprise isn't one anymore!
Guy Martin (1) explains that Amazon has a feature to avoid precisely such moments. He can remove items from his Amazon purchase history. Now the secret gift will not be used for recommendations.
This feature can also help us make healthier choices. For example, you can remove past cookie purchases from your history. A bit like someone who removes all tempting cookies from home when on a diet. Nobody wants to have someone knocking on the door to remind you of your past unhealthy habits.
So let me ask you:
What are valuable recommendations you can give automatically to people? How can you let people customize these recommendations (2)?
Footnotes
A big thank you to Guy Martin for sharing the service experience that inspired this principle with the co-creator community.
(1) A smart shopper and a member of the co-creator Service Design community
(2) On Spotify, you can love or dissmiss songs to improve the recommendations. On Youtube, you can improve your recommendations with a feedback button. On Netflix, you can create different profiles. And maybe create one only for your guilty pleasures. To explore why profiles are a smart idea, read the principle “Remember that I have multiple roles in my life.” — Daniele Catalanotto (2022). Remember that I’ll share my account. Swiss Innovation Academy. Available at https://extra.swissinnovation.academy/Pwf6 accessed.
Daniele’s notes
- This is the third draft of this principle.
- I’ve reduced the length of this principle by 51 % compared to the previous draft.