What do you think?
Rate this book
The MIT Press Essential Knowledge
Recommendation Engines
How companies like Amazon and Netflix know what “you might also like”: the history, technology, business, and social impact of online recommendation engines. Increasingly, our technologies are giving us better, faster, smarter, and more personal advice than our own families and best friends. Amazon already knows what kind of books and household goods you like and is more than eager to recommend more; YouTube and TikTok always have another video lined up to show you; Netflix has crunched the numbers of your viewing habits to suggest whole genres that you would enjoy. In this volume in the MIT Press's Essential Knowledge series, innovation expert Michael Schrage explains the origins, technologies, business applications, and increasing societal impact of recommendation engines, the systems that allow companies worldwide to know what products, services, and experiences “you might also like.” Schrage offers a history of recommendation that reaches back to antiquity's oracles and astrologers; recounts the academic origins and commercial evolution of recommendation engines; explains how these systems work, discussing key mathematical insights, including the impact of machine learning and deep learning algorithms; and highlights user experience design challenges. He offers brief but incisive case studies of the digital music service Spotify; ByteDance, the owner of TikTok; and the online personal stylist Stitch Fix. Finally, Schrage considers the future of technological Will they leave us disappointed and dependent—or will they help us discover the world and ourselves in novel and serendipitous ways?
296 pages, Paperback
Published September 1, 2020
58 people are currently reading
371 people want to read
About the author
Michael Schrage
16 books17 followersMichael Schrage is a Research Fellow at the MIT Sloan School of Management's Initiative on the Digital Economy. A sought-after expert on innovation, metrics, and network effects, he is the author of Who Do You Want Your Customers to Become?, The Innovator's Hypothesis: How Cheap Experiments Are Worth More than Good Ideas (MIT Press), and other books.
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 19 of 19 reviews
December 26, 2020
I found this a very one of a kind read I was in great need of. I work on recommendation systems.
This book sets in perspective the history and greater role of recommendations in human life. It marries the ideas of general advice seeking to providing specific recommendations in say, an app such as spotify/youtube. You get to understand the history of advice seeking - for eg folk wisdom can be thought of as one kind of advice seeking; so too divination and also interestingly astrology. Recommendation or advice seeking did not start out as a personalization or optimization problem, much like the history of medication where people initially listened to quacks (even as recently until the Spanish Flu of the 1920s which probably added to the devastation) - and so math, Machine Learning and Artifical Intelligence are but tools (albeit powerful) to a larger problem.
We then see an overview of the technical history of rec systems as we know them today, and then an overview on how they work for the semi technical reader.
The following chapter speaks to how users experience recommendations, which is useful; it has actionable suggestions based on insights from numerous domains including behavioural economics to improve recommendation outcomes - eg social validation measures where you notify people that other "awesome" people are doing this or that thing, and trust buiding methods where you explain why you recommend things can be crucial.
This is followed by a case study of Spotify, Stitch Fix and Byte Dance. This is a very exciting peek into the world of these companies and what they have done to differentiate themselves. An overview of the Spotify recommendation system was very helpful; so too insights into the methods of Byte Dance which highlight how the product is as much recommendation as platform. Learning the Stitch Fix actively has humans at the end of the recommendation loop was something new I learnt.
The final chapter speaks to an ambitious future about multiple selves being constantly improved on by using recommendations. This divorces the chapter from the specifity of say music from the rest of the book and is as such separate in content. I am not a fan of prophesying the future, but the idea that these mechanisms will holistically tie into our lives is what I take back.
An overarching idea I take back from this book is that as much as a technical problem, recommendation is a choice problem; what we know about how humans make decisions is the first source of information; and only then do technical methods help. As such the ideal recommender will leverage psychology, economics and then math/AI/computer science.
This book sets in perspective the history and greater role of recommendations in human life. It marries the ideas of general advice seeking to providing specific recommendations in say, an app such as spotify/youtube. You get to understand the history of advice seeking - for eg folk wisdom can be thought of as one kind of advice seeking; so too divination and also interestingly astrology. Recommendation or advice seeking did not start out as a personalization or optimization problem, much like the history of medication where people initially listened to quacks (even as recently until the Spanish Flu of the 1920s which probably added to the devastation) - and so math, Machine Learning and Artifical Intelligence are but tools (albeit powerful) to a larger problem.
We then see an overview of the technical history of rec systems as we know them today, and then an overview on how they work for the semi technical reader.
The following chapter speaks to how users experience recommendations, which is useful; it has actionable suggestions based on insights from numerous domains including behavioural economics to improve recommendation outcomes - eg social validation measures where you notify people that other "awesome" people are doing this or that thing, and trust buiding methods where you explain why you recommend things can be crucial.
This is followed by a case study of Spotify, Stitch Fix and Byte Dance. This is a very exciting peek into the world of these companies and what they have done to differentiate themselves. An overview of the Spotify recommendation system was very helpful; so too insights into the methods of Byte Dance which highlight how the product is as much recommendation as platform. Learning the Stitch Fix actively has humans at the end of the recommendation loop was something new I learnt.
The final chapter speaks to an ambitious future about multiple selves being constantly improved on by using recommendations. This divorces the chapter from the specifity of say music from the rest of the book and is as such separate in content. I am not a fan of prophesying the future, but the idea that these mechanisms will holistically tie into our lives is what I take back.
An overarching idea I take back from this book is that as much as a technical problem, recommendation is a choice problem; what we know about how humans make decisions is the first source of information; and only then do technical methods help. As such the ideal recommender will leverage psychology, economics and then math/AI/computer science.
September 26, 2021
Schrage helps clarify how omnipresent recommendation AI is, and how quickly it's learning from us and personalizing the world to each one of us. It's very strange to read someone so fully aware of this technology's ethical problems, and yet so myopically optimistic about it.
Can YouTube, Amazon, Facebook, ByteDance, and the others manipulate us by recommending an impulse buy, a 53rd video to binge, or another post to reply to in outrage? Will these recommendation systems end up throwing us under the bus for the sake of maximizing engagement, revenue, or any other preferred metric of success? Schrage ultimately dismisses these concerns arguing that (1) recommenders que users feel satisfied will ultimately be more successful [he offers no evidence for this though] and (2) recommenders that use the tools of behavioral science to manipulate instead of serve its users are just not really recommendation engines. Neither of these two arguments seem to promise a bright future for users.
Despite its optimistic tone, this book leaves me rather worried. The last chapter (on the future of the self and individual agency in the age of AI) is a striking case of unintentional dystopic sci-fi. I particularly recommend it.
Can YouTube, Amazon, Facebook, ByteDance, and the others manipulate us by recommending an impulse buy, a 53rd video to binge, or another post to reply to in outrage? Will these recommendation systems end up throwing us under the bus for the sake of maximizing engagement, revenue, or any other preferred metric of success? Schrage ultimately dismisses these concerns arguing that (1) recommenders que users feel satisfied will ultimately be more successful [he offers no evidence for this though] and (2) recommenders that use the tools of behavioral science to manipulate instead of serve its users are just not really recommendation engines. Neither of these two arguments seem to promise a bright future for users.
Despite its optimistic tone, this book leaves me rather worried. The last chapter (on the future of the self and individual agency in the age of AI) is a striking case of unintentional dystopic sci-fi. I particularly recommend it.
February 10, 2022
This book talked more about things unrelated to recommendation engines than recommenders themselves. A more accurate title for this book is "a very long philosophy essay and ponderings of recommendation's influence on society."
January 19, 2025
When we give advice, what is it that we are giving? A worldview, a bid for relationship, novelty, surprise, or serendipity?
This book features a history of the process of advice giving, and how it morphed into the science of being irresistible.
Chapter 1 defines web 2.0, “collective use improves data“, and chapters two and three feature history digressions for the most part.
This subject is best understood by case study. Chapter 3 on the Netflix prize and the three case studies in chapter 6 are the best ways to quickly and easily understand what fascinates you about the subject.
I am personally not particularly interested in history or in the choice architecture involved in their deployment, but more in the fact that advice has become automated now. What does this mean for us?
For example, challenge you to listen to your Spotify Discover weekly 30 songs and see if you love half of them and like 70% of them as predicted?
If you don’t have much time chapter 6 with the three case studies and chapter 3 on Netflix are the best places to start.
I benefited from the reference to Pedro Domingo’s essay on a few useful things to know about machine learning. https://sites.astro.caltech.edu/~geor...
The further reading section is elegantly limited to only 10 suggested readings. I appreciate that. I recommend working with AI however now is read to the books.
An excellent and thought provoking read thank you!
This book features a history of the process of advice giving, and how it morphed into the science of being irresistible.
Chapter 1 defines web 2.0, “collective use improves data“, and chapters two and three feature history digressions for the most part.
This subject is best understood by case study. Chapter 3 on the Netflix prize and the three case studies in chapter 6 are the best ways to quickly and easily understand what fascinates you about the subject.
I am personally not particularly interested in history or in the choice architecture involved in their deployment, but more in the fact that advice has become automated now. What does this mean for us?
For example, challenge you to listen to your Spotify Discover weekly 30 songs and see if you love half of them and like 70% of them as predicted?
If you don’t have much time chapter 6 with the three case studies and chapter 3 on Netflix are the best places to start.
I benefited from the reference to Pedro Domingo’s essay on a few useful things to know about machine learning. https://sites.astro.caltech.edu/~geor...
The further reading section is elegantly limited to only 10 suggested readings. I appreciate that. I recommend working with AI however now is read to the books.
An excellent and thought provoking read thank you!
March 14, 2024
Good, solid intro to the history of recommendation systems, their original story, different types, and what the future of recommenders could look like. Best used in conjunction with other learning materials such as a course, since it has more details into the philosophy and history of recommenders than I expected. Don’t expect to learn how to implement recommenders, but read it if you’re generally interested in the high level applications of recommenders. Good weekend read, would recommend!
November 14, 2020
Surface level only. Wanted more high level implementation details in an introductory text.
March 11, 2023
The state of recommendation engines is chaotic. Much of the book deals with Netflix's use model, while the remainder deals with user reviews as the basis of recommendations. Add in a bit of positive-thinking nonsense, and propaganda theory. Too often the answer was a digital echo chamber.
“Their “deceptively simple idea,” as they observed in a 1994 paper, was that people who had agreed in the past were more likely to agree in the future.”
TikTok is wildly successful in part because of its data mining. They have a vast recommendation engine underpinning the app. Like all apps, increasing user engagement is key. Even the Chinese government is concerned with the amount of data collected.
Stitch-Fix also has a huge analysis engine at the core of its business.
The book clearly shows that huge investments in analytics and AI are going to continue to be made (not unlike an arms war in computation applied to recommendations).
The last chapter tries to walk a fine line between influencing and individual choice. Based on the current state of the world, that line has clearly crossed over mega influencing. It was published at the start of the covid pandemic, and since we have learned that user engagement at all cost is front and center, truth often ignored since it may not encourage engagement.
“Their “deceptively simple idea,” as they observed in a 1994 paper, was that people who had agreed in the past were more likely to agree in the future.”
TikTok is wildly successful in part because of its data mining. They have a vast recommendation engine underpinning the app. Like all apps, increasing user engagement is key. Even the Chinese government is concerned with the amount of data collected.
Stitch-Fix also has a huge analysis engine at the core of its business.
The book clearly shows that huge investments in analytics and AI are going to continue to be made (not unlike an arms war in computation applied to recommendations).
The last chapter tries to walk a fine line between influencing and individual choice. Based on the current state of the world, that line has clearly crossed over mega influencing. It was published at the start of the covid pandemic, and since we have learned that user engagement at all cost is front and center, truth often ignored since it may not encourage engagement.
June 24, 2022
Disappointed by the way Schrage quickly dismissed a contemporary's cautions against becoming too reliant on algorithms. I'm a pessimist, I guess -- I don't believe that everyday people, me included, utilize recommendation algorithms in order to 'become our best selves', as Schrage insists we do and will continue doing. I think it's much easier to let our lives and our efforts shrink down to whatever is enabled by the algorithm. We're lazy with our tech, and we need to be conscious of how our lives are shaped by that. Overall, the book is a little too technical in places (e.g. when he starts talking about different types of algorithms as if everyone thinks easily in terms of matrices). It seems more like a book whipped up quickly to buff up the MIT 'Essential Knowledge' series than to truly educate. I come away with it mostly with a sour taste in my mouth for all of the digital services around me that I now know are tracking every minute action -- and lack of action -- I take while browsing. It makes me long for Blockbuster.
May 6, 2021
A fascinating look into the tech that is literally shaping the world around us faster than we understand. Great survey that was information rich without leaving the layperson behind. Some did feel like recommendation engine apologetics, but they say whatever side a person researches for a debate is the flavor cool aide they inevitably drink. Possibly a 5 star book, but the author went and called Work Rules! an excellent publication - and speaking of cool aide, that was a full on overrated cool aide fest.
January 6, 2022
This book was downright depressing. The state of the world is insane and something needs to be done. Not until the start of the last chapter did the author say anything about this! He dedicated about 5 pages toward this being a problem...
The book was well-written and I did learn a ton about applied ML. We all need to work together to make sure that recommendations are strictly presented as options. TikTok is the prime example of recommendations being presented one by one. Scary stuff.
The book was well-written and I did learn a ton about applied ML. We all need to work together to make sure that recommendations are strictly presented as options. TikTok is the prime example of recommendations being presented one by one. Scary stuff.
September 15, 2023
This is an excellent non-technical view of recommendation engines and how they've evolved as well as may progress for the future. I heard a podcast the author was on in predicting we're due for a versioning multiplicity of ourselves and Im here for it. As companies proliferate more data, there'd be a responsibility to suggest improvements. Tides will rise all boats in the AI-laden decade forthcoming.
October 23, 2021
The book is good at the surface level to understand the history of recommendation engines and the real-world use cases. Many parts of the book felt redundant to me. I was more fascinated with the chapter about how big companies like Netflix, Stitch Fix, Tiktok, etc. perceive and design their recommendation engines.
May 1, 2023
Gratamente sorprendido. Creo que esta bastante bien como primera introducción a los sistemas de recomendación. Si tienes un brackground solido en el tema no creo que aprendas muchas cosas nuevas.
Recomendaría como libro introductorio a estudiantes
OOOOOO
tambien te puedes ver mi charla en el DesgranandoCiencia9 >:)
Recomendaría como libro introductorio a estudiantes
OOOOOO
tambien te puedes ver mi charla en el DesgranandoCiencia9 >:)
October 21, 2023
I will say it was a good introduction of recommendation engines, but I feel like the book could have gone farther in some areas. Also the author is a bit biased (in my opinion) towards leaning heavily on recommendation engines, as if they are the end all be all of society.
June 19, 2020
Learned a ton about recommendation engines and had my eyes opened to the influence they have over us in so many areas of our lives. Read the galley; pub date: Sept 2020
June 17, 2021
DNF. Surface level book, just touching the basic aspects of recommendation systems.
August 3, 2021
Great book on recommendation engines and some unique perspective about the history of advice and people selling advice to become better versions of themselves.
July 9, 2024
Very good book about main algorithms used in the internet until LLMs
December 14, 2021
Fantastic book. Enjoy the the entire content, love the written style of the author, brilliant. Chapter 2 'the origin', somehow too philosophical to understand for me. Chapter 4, 6 are amazing. (It doesn't go too deep about implementation, but at least can guide you where you can dive in furthur, that is more than enough for a essential series). Chapter 7 is thought-provoking (the multiselves term is interesting). Appreciate.
Displaying 1 - 19 of 19 reviews