Latest / The Adults In The AI Room / Michael Schrage on Recommendation Engines - Episode #14

Transcript

Michael Schrage on Recommendation Engines - Episode #14 · The Adults In The AI Room

• 0:22
  SPEAKER_00: Thanks so much for listening.
• 0:23
  SPEAKER_00: If this is your first time we want you to subscribe, we're going to be listening to the podcast.
• 0:29
  SPEAKER_00: You're your host.
• 0:30
  SPEAKER_00: Seth Early and Chris Featherstone.
• 0:36
  SPEAKER_02: Well, welcome to today's podcast.
• 0:39
  SPEAKER_02: I am Seth Early.
• 0:40
  SPEAKER_05: And I'm Chris Featherstone.
• 0:41
  SPEAKER_05: It's good to be with you.
• 0:42
  SPEAKER_02: So nice to uh catch up with you, Chris.
• 0:45
  SPEAKER_02: I know you may be in Boston next week.
• 0:47
  SPEAKER_02: So um hoping uh hoping we can get together and uh have you referred fire.
• 0:51
  SPEAKER_05: Yeah, we will well we should be doing these in you know more in person.
• 0:54
  SPEAKER_05: And had I known that you know Michael's in Michael was in New York, right?
• 0:57
  SPEAKER_05: I'll you'll introduce him in a minute, but I would have got together with him there.
• 1:00
  SPEAKER_05: I was there last week, all week.
• 1:01
  SPEAKER_05: There you go.
• 1:02
  SPEAKER_05: It was kind of crazy though to see like uh I'd never been in New York City for St.
• 1:06
  SPEAKER_05: Patty's Day, and it was an interesting experience, right?
• 1:10
  SPEAKER_05: Because the only hotel that I was able to stay at, which we're we're there for a present thing, was right in Times Square, which was you know not ideal.
• 1:18
  SPEAKER_02: You're lucky it wasn't Boston on St.
• 1:20
  SPEAKER_02: Patrick's Day.
• 1:21
  SPEAKER_02: I wish it was.
• 1:22
  SPEAKER_02: I wish it was.
• 1:23
  SPEAKER_02: A lot of drunk people and uh dreams.
• 1:26
  SPEAKER_05: Well, I apparently um there is this thing that I that I got to witness called Irish yoga, which apparently happens after the binge and everybody's just folded in half, uh, you know, in different parts of the wherever they land, right?
• 1:39
  SPEAKER_03: Well, you'll you'll forgive my interjection here, but I of course retain a fondness for Irish Alzheimer's where you forget everything except the grudges.
• 1:48
  SPEAKER_03: So that's awesome.
• 1:50
  SPEAKER_02: Well, let me go ahead and introduce you then, Mike.
• 1:53
  SPEAKER_02: Um it's great to have you.
• 1:54
  SPEAKER_02: Our guest today is a research fellow at the MIT Sloan School for Management and the Initiative on the Digital Economy, and he's a sought-after expert on innovation, design, and network perfect.
• 2:06
  SPEAKER_02: His book, Recommendation Engines, is a fascinating look at the history, technology, business, and social impact of online recommendation engines that have been made ubiquitous by the likes of Amazon, Spotify, uh, Netflix.
• 2:19
  SPEAKER_02: Please welcome Michael Freegi.
• 2:21
  SPEAKER_02: How are you doing?
• 2:23
  SPEAKER_03: I'm doing fairly well, and I've been looking forward to this and and much appreciate the opportunity not just to talk about quote unquote the book per se, but really the things, and this is, you know, uh, you know this, Seth, from your own experience writing a book.
• 2:37
  SPEAKER_03: You write the book and you publish it, but the real learning comes from how people interpret things or misinterpret things or engage with you and say, gee, what about so-and-so?
• 2:48
  SPEAKER_03: And and so for me, one of the great things about doing this book, outside of the relief of having done a book, uh, is is the kind of follow-ups and questions and engagement.
• 3:00
  SPEAKER_02: Right.
• 3:01
  SPEAKER_02: All of the uh conversations that it that it catalyzes.
• 3:04
  SPEAKER_02: Maybe we could maybe you could kind of back up and give us a little bit of the background of you know how you got to where you are and how you ended up at MIT.
• 3:12
  SPEAKER_02: I know you we talked about that last time.
• 3:14
  SPEAKER_02: It's a fascinating uh journey, you know, coming from being a journalist.
• 3:19
  SPEAKER_02: And uh and then we'd like to learn about how you came to writing the book.
• 3:23
  SPEAKER_02: So maybe you can give us a little bit of uh your history, the history according to Michael.
• 3:27
  SPEAKER_03: Yeah, I'm I'm I'm always uh I don't want to say shy, but but reluctant or reticent about going into aspects of my background.
• 3:36
  SPEAKER_03: Most of my behavior you're checkered, your checkered past.
• 3:40
  SPEAKER_03: It is it is it is a checkered past.
• 3:42
  SPEAKER_03: Then again, the sort of neighborhood I grew up in was more chess past or mahjong past or or go past.
• 3:51
  SPEAKER_03: Um, as as as you will now understand by I grew up as a faculty brat in Chicago's Hyde Park, and that had a big impact on my behavior and my curiosity and my approach and attitude towards learning in school.
• 4:06
  SPEAKER_03: So it was, you know, the University of Chicago Hyde Park area during the 60s, which was which was um interesting, and I learned a lot in every meaning of the phrase.
• 4:17
  SPEAKER_03: But how I got to the media lab was a little different.
• 4:21
  SPEAKER_03: Um, you know, I I after I got out of school and I was the areas that I studied were computer science and economics.
• 4:30
  SPEAKER_03: After I got out of school, I really had no interest in graduate school whatsoever.
• 4:36
  SPEAKER_03: Um, I did not want to go to school anymore.
• 4:40
  SPEAKER_03: Uh so as all ambitious uh Midwesterners do, you pick one coast or the other.
• 4:47
  SPEAKER_03: And reflecting on my my uh checkered past, my my driver's license had been taken away.
• 4:54
  SPEAKER_03: Uh, not because I drove drunk, but I like to drive too fast on the Dan Ryan.
• 4:59
  SPEAKER_03: And those, you know, when I was 16 years old, I looked like I was 12.
• 5:04
  SPEAKER_03: So I was the poor sucker who got pulled over as opposed to all the people driving 80 miles an hour, you know, who had real jobs.
• 5:12
  SPEAKER_02: Um did you have a red car?
• 5:15
  SPEAKER_02: Red cars.
• 5:16
  SPEAKER_03: No, it was it was a blue car.
• 5:17
  SPEAKER_03: It was a blue, it was it was it was a it was a Honda, for God's sake.
• 5:21
  SPEAKER_03: I mean, it was my parents' car, it wasn't my car, you know.
• 5:26
  SPEAKER_03: Um so I ended up in New York and began doing something that one can no longer do.
• 5:33
  SPEAKER_03: Uh, I was a freelance writer, and in those days, um you really could make a living freelance writing.
• 5:42
  SPEAKER_03: And because I had a technology background, uh I worked actually in high school for the Ittty Bitty Machine Company, small IBM.
• 5:51
  SPEAKER_03: It was an a store that sold Altaires and MSI computers.
• 5:57
  SPEAKER_03: I should say I one other thing about my background is what got me into computers from the get-go was a guy by the name, and I imagine you know him, Seth, and possibly you as well, Chris, uh Ted Nelson, who wrote Computer Lib, Dream Machines and Computer Lib.
• 6:13
  SPEAKER_03: And and just to put a little footnote on that, my dad actually gave Ted office space so he could finish the book.
• 6:22
  SPEAKER_03: And so Ted sort of had to be nice to me on on this regard.
• 6:27
  SPEAKER_03: He was the one who got me into the Itty Bitty Machine Company, which of course went bankrupt.
• 6:32
  SPEAKER_03: But but anyway, uh, so I'm in New York and I'm writing and covering technology, and I end up being the first video columnist, slash technology columnist, slash personal computer columnist for Rolling Stone magazine, Jan Wenner.
• 6:51
  SPEAKER_03: Yeah, and Jerry McDonald and Dave Rosenthal were the editors.
• 6:55
  SPEAKER_03: And it was funny because the advertising people loved me.
• 6:58
  SPEAKER_03: By the way, I could write about anything I wanted having to do with technology, except um, you know, records, except stereo equipment, because they were the other columnists for that.
• 7:09
  SPEAKER_03: Yeah, but I was the person.
• 7:13
  SPEAKER_03: Exactly.
• 7:13
  SPEAKER_03: Exactly.
• 7:14
  SPEAKER_03: I was writing about video games, I was writing about VCRs, I was writing about DVDs, you know, remember the video disc, you know, that kind of stuff.
• 7:26
  SPEAKER_03: And I so it got me going to the consumer electronic show.
• 7:31
  SPEAKER_03: And uh long story short, um, the Washington Post became interested in my stuff.
• 7:38
  SPEAKER_03: And I ended up writing for the business desk at the Washington Post.
• 7:42
  SPEAKER_03: And that was a great, great experience.
• 7:45
  SPEAKER_03: And for me, it was sort of like grad school because I was always writing papers.
• 7:49
  SPEAKER_03: I got to talk to everybody.
• 7:50
  SPEAKER_03: The great thing about the paper at that time, we won't talk about the newspaper these days, but you know, you could have access to everyone.
• 7:58
  SPEAKER_03: And the other great thing about it being the Post versus the Times or the Wall Street Journal is that the Washington Post didn't pretend to try to be comprehensive about business or technology.
• 8:09
  SPEAKER_03: It just wanted to be able to offer, you know, the Washington elite, the Washington literati insight into these sort of things.
• 8:18
  SPEAKER_03: So I really could indulge that extra day of reporting and coverage.
• 8:23
  SPEAKER_03: Uh I I was I wasn't a stupid guy, so I also covered defense technology, the you know, GPS, Brad Parkinson.
• 8:32
  SPEAKER_03: So I had a pretty broad set of takes on on that.
• 8:38
  SPEAKER_03: And I worked there.
• 8:39
  SPEAKER_03: And sorry for being so long-winded, but one of the I ended up coming back to New York to be the financial correspondence, to be the financial correspondent.
• 8:49
  SPEAKER_03: And and they wanted me to come back to Washington to be promoted to stuff, but I didn't want to go back to Washington.
• 8:56
  SPEAKER_03: I had had a lovely, lovely girlfriend there, and we had broken up, and I didn't really want to go back.
• 9:02
  SPEAKER_03: And and I just, you know, wanted to do something else.
• 9:05
  SPEAKER_03: I also didn't see myself writing like David Broder for the next 30 or 40 years.
• 9:12
  SPEAKER_03: And and um one of the companies was company, shame on me.
• 9:16
  SPEAKER_03: One of the stories I did when I was at the Washington Post was on this crazy, crazy startup at MIT called the Media Lab, but it wasn't really the Media Lab.
• 9:26
  SPEAKER_03: In those days, it was the Architecture Machine Group or ArcMac.
• 9:30
  SPEAKER_03: And, you know, this crazy guy, Nick Negraponte, which by the way, Nicholas Negraponte's brother was John Negraponte, who ran the Contras in Honduras, and he was a big Washington macher, defense and state department mocher.
• 9:50
  SPEAKER_03: And he actually think he was the head of the intelligence community for a while.
• 9:54
  SPEAKER_03: So, so turns out the post that kind of stuff was interesting.
• 9:58
  SPEAKER_03: So these weird serendipitous connections, but it's too late for me to say long story short.
• 10:03
  SPEAKER_03: But when I told that I was interested in doing a book on collaboration and collaborative technologies, just like my father had done for Ted Nelson, Nicholas Negroponte did for me, he invited me to be a fellow at the Media Lab and finished my book.
• 10:20
  SPEAKER_03: And by the way, do you know who the previous fellow at the Media Lab who finished a book there was?
• 10:27
  SPEAKER_03: Who was Stuart Brand, who wrote a book called The Media Lab.
• 10:31
  SPEAKER_03: So from Holer Catalog to Media Lab.
• 10:34
  SPEAKER_03: So that those were the two big shoes for me to fill.
• 10:38
  SPEAKER_03: But I fell in love, I fell into MIT, fell in love, and I've sort of had a relationship dotted and solid line for decades.
• 10:48
  SPEAKER_02: And I'm just sorry you asked.
• 10:52
  SPEAKER_02: Sorry?
• 10:53
  SPEAKER_02: What was what what year was that when you first came to MIT?
• 10:58
  SPEAKER_02: Um I'd say 85, 86.
• 11:01
  SPEAKER_02: Wow.
• 11:02
  SPEAKER_03: Wow.
• 11:02
  SPEAKER_03: Okay.
• 11:03
  SPEAKER_03: Cool.
• 11:03
  SPEAKER_03: Yeah.
• 11:04
  SPEAKER_03: I was young.
• 11:04
  SPEAKER_03: I was I was very young when I was I used to there was a time when I was the youngest guy in the room, the youngest person in the room.
• 11:11
  SPEAKER_03: And and those days are just gone.
• 11:13
  SPEAKER_03: Those days are gone.
• 11:17
  SPEAKER_03: The only time I now am the youngest person in the room is when I go to retirement parties and funerals.
• 11:27
  SPEAKER_02: Yeah, I'm I'm not the youngest person in the room when I go to anything.
• 11:33
  SPEAKER_02: But with the exception of what you just said, yeah.
• 11:35
  SPEAKER_05: Well, I mean it wouldn't feel as so um as so geriatric as it does today because technology advances so quickly, right, at the same time.
• 11:46
  SPEAKER_05: So we're dealing with with multiple timelines of our own time and our own space, but then you also have these other things that are accelerating past us at the same time that we're trying to keep, you know, stay ahead of or on top of.
• 12:00
  SPEAKER_05: And it's tough to have all the latency of the of you know of yesteryear to actually get up speed on it quick because we kind of made things easier.
• 12:10
  SPEAKER_02: I don't know if that was on my side or your side, Chris.
• 12:13
  SPEAKER_05: Um it was just it was just the the you know, that that thought of the technology the technology speed in our own time in uh in life, you know, feels like there's there can be sometimes, you know, passing each other in the night, and then all of a sudden you wake up and lo and behold, you know, there's a whole new something, right?
• 12:33
  SPEAKER_05: That oh yeah you need to get ahead of.
• 12:35
  SPEAKER_03: So well, I I gotta say, you know, just building on what your your your point, I gotta say that that one of the things that I've always felt in technology, covering technology, is overwhelmed.
• 12:49
  SPEAKER_03: There's so much going on.
• 12:52
  SPEAKER_03: You don't really know ex ante what's going to win or what's going to lose.
• 12:58
  SPEAKER_03: I again, one of the biggest pieces I wrote 40 years ago was on video discs and where are they now?
• 13:05
  SPEAKER_03: You know, they're like they're like LP records, and somehow I don't think they're coming back.
• 13:10
  SPEAKER_02: They're coming back.
• 13:10
  SPEAKER_02: I love that line from 30 talk.
• 13:12
  SPEAKER_02: The beeper king technology is psychological.
• 13:15
  SPEAKER_03: Yeah, yeah, the beeper king.
• 13:18
  SPEAKER_03: I I but it's it's you know, and and I'm I may or may not use this as a segue or a bridge into the book or other things, but one of the most important things I came away with from that feeling of being overwhelmed is if there's one thing that'll get you through overwhelming change, it's get some sort of purchase, get some sort of grasp, get some sort of actionable insight into what the fundamentals really are.
• 13:48
  SPEAKER_03: What are the real fundamentals of the of the technology?
• 13:52
  SPEAKER_03: You know, what does being digital really mean?
• 13:55
  SPEAKER_03: What what is what does it mean to come?
• 13:59
  SPEAKER_03: What is software?
• 14:01
  SPEAKER_03: What does what does compiling versus interpreting mean?
• 14:04
  SPEAKER_03: What what's what's the essence that matters and that endures?
• 14:09
  SPEAKER_03: So that even if you're introduced to something you have no bloody clue about, you can ask a question that at least makes you appear smart.
• 14:18
  SPEAKER_03: And and now I'm I'm revealing the great secret.
• 14:21
  SPEAKER_03: You know, it's very, very important for quote unquote journalists and reporters to appear smart.
• 14:27
  SPEAKER_03: The only profession that needs professions that are more compelled to appear smart are lawyers and professors.
• 14:37
  SPEAKER_05: Hey, Michael, let me let me ask you something.
• 14:39
  SPEAKER_05: Um given given that background, when did you start to peer into like the behavioral economics of of such innovation and stuff?
• 14:48
  SPEAKER_05: Because I know that's a key thing that you that you like to write and advise about.
• 14:52
  SPEAKER_05: And then that will lead into, let's say, recommendations, right?
• 14:55
  SPEAKER_05: Because they are, you know, are so tightly coupled.
• 14:59
  SPEAKER_05: But when did you start to notice like just the behavioral pieces and what that could do to drive the you know economics?
• 15:04
  SPEAKER_05: And then you know, speak about that a little bit.
• 15:07
  SPEAKER_03: Well, that is a great question.
• 15:09
  SPEAKER_03: And and I'm I'm gonna get gonna offer an answer that that you both of you and whoever chooses to listen to this may not believe.
• 15:18
  SPEAKER_03: But um, I actually, so my dad was a scientist.
• 15:23
  SPEAKER_03: We uh got science, we got the Chronicle of Higher Education and Science Magazine, both of which I read, I won't say religiously, but there was a there was an article by Kahnman and Tversky in Science magazine.
• 15:37
  SPEAKER_03: And you know, I couldn't understand the chemistry and the physics, but I understood this one on judgment under uncertainty and the notion of cognitive biases and prospect theory.
• 15:47
  SPEAKER_03: And this struck me, even in high school, it struck me because I was always interested in how do people make decisions?
• 15:53
  SPEAKER_03: Why?
• 15:54
  SPEAKER_03: Why do you do this?
• 15:55
  SPEAKER_03: Okay, so one of my majors in school is economics.
• 16:00
  SPEAKER_03: All the economists talk when you when you take a standard curriculum in in economics, it's about what?
• 16:06
  SPEAKER_03: Not positive economics, you know, data-driven.
• 16:09
  SPEAKER_03: This is like experimental, you know, so much for Vern Smith and experimental experimental economics.
• 16:15
  SPEAKER_03: In economics, it's you assume that that people are trying to make the right decision, the optimal decision.
• 16:22
  SPEAKER_03: It it most of economics behaves as if psychology doesn't exist.
• 16:28
  SPEAKER_03: Now, you may recall one of the guys who won a Nobel Prize, but who was also one of the pioneers in AI and and um you know artificial intelligence and decision theory.
• 16:42
  SPEAKER_03: And that was Herb Simon, who won the Nobel Prize.
• 16:45
  SPEAKER_03: And he spoke at Illinois.
• 16:47
  SPEAKER_03: I got to talk to Herb Simon, and Herb Simon was one of the people, you know, yeah, he was interested in technology, software AI.
• 16:55
  SPEAKER_03: Uh, Simon and oh, come on, what's his name?
• 16:58
  SPEAKER_03: Newell and Simon, Alan and New Ellen Newell and Herb Simon.
• 17:02
  SPEAKER_03: But he he'd also introduced the concept of what in decision-making and economics?
• 17:07
  SPEAKER_03: Satisficing.
• 17:08
  SPEAKER_03: Economics, they talk about maximizing.
• 17:11
  SPEAKER_03: But the point that Simon made about bounded rationality, limits on your ability to compute and indeed survey the territory, meant that you were trying to satisfice, you wanted to have good enough decisions, because you didn't know enough to be able to make the best decision.
• 17:30
  SPEAKER_03: Yeah.
• 17:30
  SPEAKER_03: So your your your uh your anticipation or your setup that this is the kind of thing that would lead to my interest or resonance with recommendation is spot on.
• 17:42
  SPEAKER_03: Because I realized, unlike some Nobel laureates, that that um we're always going to have gaps in knowledge, gaps in in capability.
• 17:54
  SPEAKER_03: And so the notion of a perfect choice, the best choice is a shimera, it's a unicorn, it's a myth.
• 18:02
  SPEAKER_03: People are biased, right?
• 18:04
  SPEAKER_03: Human beings, we're flawed.
• 18:07
  SPEAKER_03: What are the right kind of choice architectures for people who are such flawed, biased, and poorly informed decision makers?
• 18:19
  SPEAKER_03: And that ironically has kind of been a theme, you know, and that that that really was a theme.
• 18:25
  SPEAKER_03: That was a sensibility that came out of college.
• 18:27
  SPEAKER_03: And forgive my one last passive-aggressive insult.
• 18:30
  SPEAKER_03: But when I went to the Washington Post and I was in the center of power and went to New York, you know, I gotta tell you, I was not blown away by our elected officials.
• 18:41
  SPEAKER_03: I was not blown away.
• 18:42
  SPEAKER_03: I mean, I was, geez, these are pretty big decisions you're making off of such crappy information.
• 18:49
  SPEAKER_03: And so this is really something that is that is, I wouldn't say obsessed me, but it's haunted me and has concerned me for as long as I've been an adult.
• 18:58
  SPEAKER_05: You know what?
• 18:59
  SPEAKER_05: No, it's it's interesting, and you could tell like those pivotal moments that you that get um like almost branded into your brain, right?
• 19:09
  SPEAKER_05: Like you were you were talking about that article that you read.
• 19:11
  SPEAKER_05: I mean, you still recall every little intricate detail about that because it was so impactful, and it probably opened up how many aha moments for you, right?
• 19:19
  SPEAKER_05: Say, oh good Lord, this is there's something important about this.
• 19:24
  SPEAKER_05: It's touching my soul in a way that I that I really want to figure out and probably, you know, talk about bias, probably gave you a good framework of thinking as you're writing about this technology, that technology, and all those kind of things, and the for you know, the the kind of the first foundations of how types of things will, you know, frame it.
• 19:44
  SPEAKER_05: So I love it.
• 19:44
  SPEAKER_05: I mean, that was yeah.
• 19:46
  SPEAKER_03: But Chris, you're spot on.
• 19:47
  SPEAKER_03: The only snap in the tail I'd put is that the Kommentversky prospect theory approach changed my notion about key fundamentals associated with decision making.
• 20:02
  SPEAKER_03: In other words, you can't assume that people can be rational maximizers.
• 20:07
  SPEAKER_03: You have to assume that they're biased and that there needs to be some sort of mechanism or approach or presentation to compensate or mitigate or ameliorate that bias.
• 20:20
  SPEAKER_03: So it's sort of like flipped my notion of what the fundamentals were, and that's a big deal.
• 20:26
  SPEAKER_05: Yeah, no, that's awesome.
• 20:27
  SPEAKER_02: So maybe you could talk about what led to the um you writing the book and then and then talk about you know the idea of recommendation versus personalization.
• 20:36
  SPEAKER_02: And uh you you did mention choice architecture, so maybe we could kind of dive in.
• 20:41
  SPEAKER_02: I know those are pretty broad areas, but um pick some of those apart.
• 20:46
  SPEAKER_03: I had done, I was doing research, I'd written a number of books, but you know, one of the things I always pay attention to, of course, uh, you know, self-absorbed narcissist that I am, is me.
• 20:58
  SPEAKER_03: You know, remember, I'm concerned about I'm concerned about decision making, I'm concerned about influence.
• 21:05
  SPEAKER_03: You know, these are these are I I I teach students, I advise clients, I'm so I'm I'm interested in the fundamentals associated with their decision making and how they uh uh perceive and act upon choice.
• 21:21
  SPEAKER_03: And I had been noticing for myself, and I actually wrote a piece in the MIT uh technology review, that that in things like music and and uh um music and travel, I was strongly, I was increasingly relying, and particularly books with Amazon, I was increasingly relying upon recommendation engines to expose me to things that I otherwise wouldn't have come across.
• 21:51
  SPEAKER_03: And this was a really intriguing thing to me because I did, I really did have this.
• 21:57
  SPEAKER_03: I I'm not gonna say it's a eureka moment.
• 22:00
  SPEAKER_03: It's sort of a sobering moment.
• 22:02
  SPEAKER_03: Um, it was a moment where I realized that actually I was getting better reading recommendations from Amazon than from my friends.
• 22:10
  SPEAKER_03: You know, love my friends.
• 22:12
  SPEAKER_03: Yeah, absolutely.
• 22:13
  SPEAKER_03: They're great, they're wonderful people, but but um actually uh some of the recommendations I was getting from Amazon and you know other recommenders was of greater use to me.
• 22:27
  SPEAKER_03: Um and so I I did a little essay on on this, and then it just I just let it simmer.
• 22:34
  SPEAKER_03: But but you know, what let's go back to the Media Lab.
• 22:37
  SPEAKER_03: Patty Maz's group did uh did a project called Firefly, which ultimately got sold to to uh Microsoft, which shut it down stupidly.
• 22:47
  SPEAKER_03: But it was a social, you know, it was like a collaborative filter, it was one of the first collaborative filters.
• 22:52
  SPEAKER_03: I was did a bit of work with uh um Tom Malone at the Sloan School, the Center for Coordination Science, and Tom is doing preliminary work on recommenders.
• 23:02
  SPEAKER_03: And I so so I had been exposed to this, and you know, full disclosure, you know, did I ever think that they would get to be as big as they were?
• 23:12
  SPEAKER_03: No.
• 23:12
  SPEAKER_03: And that was actually one of the reasons why I wrote the book, because you know, I know the editor, the editorial director of of uh the MIT Press, Gidemanaktala, you know, lovely, smart, terrific uh person and editor.
• 23:29
  SPEAKER_03: She had published one of my books, the book on experimentation earlier, and they were doing a series on essential knowledge, and it wouldn't take up too much time, ha ha ha.
• 23:39
  SPEAKER_03: And so he said, write about something that you really care about, that that you, Michael, I know that you care about the fundamentals, you know, something that you care about and you think is going to matter more, but but something that you relate to personally, but you think is also a big deal professionally.
• 23:54
  SPEAKER_03: And I thought, you know, I think I should do recommendation engines.
• 23:59
  SPEAKER_03: And and you know, I knew the technology well enough to bluff, but I didn't really know the technology.
• 24:07
  SPEAKER_03: And so I basically used the book as an excuse.
• 24:12
  SPEAKER_02: All of us only know enough about technology to bluff at some level.
• 24:18
  SPEAKER_03: Yeah.
• 24:19
  SPEAKER_03: Um, I uh well, you know, it depends on the the hand that you're dealt about how much the bluff matters, but um it was fascinating.
• 24:28
  SPEAKER_03: Now, the one wild card in all of this, which I think is not irrelevant to this conversation and to anybody who would choose to listen to this, is as I wrote the book, something happened during the period of time writing the book that I did not expect to happen.
• 24:48
  SPEAKER_03: I expected it to happen later.
• 24:50
  SPEAKER_03: And what was that?
• 24:51
  SPEAKER_03: Machine learning began to colonize recommenders.
• 24:58
  SPEAKER_03: It wasn't just how do we come up with better market basket rules, you know, better uh correlation, that kind of it basically how do we learn?
• 25:09
  SPEAKER_03: How do we learn?
• 25:11
  SPEAKER_03: And and the the recommended recommendation engines became learning to recommend engines.
• 25:19
  SPEAKER_03: And one of the things that I did in the book that that actually a couple of the peer reviewers rejected, but but fortunately my editor and I agreed belonged in the book, is that we realized that if you took recommendation engines seriously, if you took the fundamentals seriously, that that this is really a book about uh advice, the technologies of advice.
• 25:46
  SPEAKER_03: Casting a horoscope, what is your what is your horoscope on an algorithm based on the stars?
• 25:53
  SPEAKER_03: Okay, same thing with with uh um uh the i ching, you know, with its its hexagrams.
• 26:03
  SPEAKER_03: I was really writing a book not about recommendation engines per se, but the nature of technology of advice.
• 26:12
  SPEAKER_03: How do we use tools and technologies to come up with better and more reliable advice?
• 26:20
  SPEAKER_03: And that was really the narrative spine, the narrative arc, and forgive the use of this word, the transcendent insight that drove the book, not recommendation engines per se, but what do we want advice to mean?
• 26:38
  SPEAKER_03: What do we want advice to be?
• 26:40
  SPEAKER_03: How can we make advice more valuable?
• 26:43
  SPEAKER_05: You know, you go into the history of it as well, right?
• 26:47
  SPEAKER_05: Not like the the this isn't this is basically an application of an old pattern applied to you know to technology, new concepts to apply that advice.
• 26:59
  SPEAKER_05: Is that fair?
• 27:00
  SPEAKER_03: Exactly right, exactly right.
• 27:03
  SPEAKER_03: But but because technology changes, the the because the semantics and the ontology and the technology and the data change, the meaning of advice, the meaning of the personalization of advice, the meaning of the presentation of advice changes.
• 27:23
  SPEAKER_03: As I mentioned in the book, and I'll stick with this right now, there is a hundred billion dollar valuation for a company that figures out how to turn AR to create it advisory augmentation reality experiences.
• 27:42
  SPEAKER_03: You put your phone or your device over something and it advises you about where to go, what to do, et cetera.
• 27:48
  SPEAKER_03: You know, yes, we understand the notion of a recommendation listing something or showing something, or you're saying pick this movie, you know, on Netflix, et cetera, or this song on Spotify, what have you.
• 28:00
  SPEAKER_03: But but changing the form factor, changing the medium enables a transformation of advice opportunity.
• 28:08
  SPEAKER_03: Let's call it the AX, the advice experience.
• 28:12
  SPEAKER_03: I think that's collectively worldwide, throughout China in, we're talking at least there's at least a trillion dollars sitting sitting around an under underrealized, underappreciated value.
• 28:27
  SPEAKER_05: Well, the the the thing that I love that you call out though is you pose the question will a recommendation engine leave you disappointed and wanting?
• 28:37
  SPEAKER_05: Or can they get to the point where it's good enough where it feels like you call out like it's serendipity?
• 28:44
  SPEAKER_05: Meaning, like, you know, instead of this weirdness that everybody's, you know, you know, hey, I've got all these hockey pucks that are listening to every goddamn thing that I'm saying on, you know, in the in the house, and then all of a sudden they're surprised when ads show up on their the mobile device or in their browser or whatever.
• 29:01
  SPEAKER_05: And it doesn't feel it feels big brotherish as opposed to serendipitous, right?
• 29:07
  SPEAKER_05: And so, you know, part of that is I think the the crux of what you're talking about too, with this is is you know, that scares people.
• 29:15
  SPEAKER_05: So does that that leave them an experience where they feel like you know they're being watched over, watched on, or or peered into, or heaven forbid, almost like a uh, you know, like you know, somebody's always looking in their window, or do they feel like, oh, somebody's actually really listening to the my need, you know, addressing my needs, not what they think I want.
• 29:39
  SPEAKER_05: Does that make sense?
• 29:40
  SPEAKER_03: Yeah, it makes sense.
• 29:41
  SPEAKER_03: I I let me let me deal.
• 29:43
  SPEAKER_03: I get asked about the ethical stuff a lot.
• 29:45
  SPEAKER_03: Let me deal with that.
• 29:46
  SPEAKER_01: We'll be right back.
• 29:47
  SPEAKER_01: We'd like to thank Marketing Artificial Intelligence Institute for their support of this podcast.
• 29:53
  SPEAKER_01: Marketing AI Institute is a media, event, and online education business that makes AI approachable, actionable, and accessible to marketing leaders around the world.
• 30:02
  SPEAKER_01: They run the world's leading marketing AI conference, MACOM, which brings together hundreds of marketing and business leaders to better understand, pilot, and adopt AI.
• 30:12
  SPEAKER_01: Today's podcast is brought to you by Simpler Media Group, owner of the Customer Experience Decoded Podcast, covering the latest in customer experience and marketing.
• 30:20
  SPEAKER_01: CX Decoded is a semi-monthly podcast designed to deliver insight from leaders driving the evolution of customer experience with an emphasis on digital touch points, customer journeys, respectful data management, and voice of the customer.
• 30:34
  SPEAKER_05: I want to give a big thanks to Early Information Science in support of the podcast.
• 30:39
  SPEAKER_05: Early Information Science is a professional service firm with proven methodologies designed specifically to address product data, content assets, customer data, and corporate knowledge bases.
• 30:48
  SPEAKER_05: Their information architecture expertise has allowed clients from all over the world to efficiently organize, access, and utilize their data, ultimately resulting in better business outcomes.
• 30:57
  SPEAKER_05: They deliver customer-driven solutions that scale and adapt to the needs of a business as it evolves.
• 31:03
  SPEAKER_05: Founded in 1994, Early Information Science has proudly delivered result-driven solutions for more than 100 brands across the top industries.
• 31:11
  SPEAKER_05: You can get more information about early information science at early.com.
• 31:16
  SPEAKER_05: E-A-R-L E Y dot com.
• 31:20
  SPEAKER_05: Working with customers in AI, the number one question that I get usually is where do I start?
• 31:26
  SPEAKER_05: What data do I need?
• 31:27
  SPEAKER_05: How do I gather it?
• 31:28
  SPEAKER_05: And it's great questions.
• 31:30
  SPEAKER_05: Well, if you want to catch the wave of AI, not be crushed by it, you really do need to read a book written by Seth Early called The AI Powered Enterprise.
• 31:38
  SPEAKER_05: It gives you the ability to look and understand how to harness the power of ontologies to make your business smarter, faster, and more profitable.
• 31:45
  SPEAKER_05: Seth goes into the intricate details of how to build an ontology, develop and employ an ontology as a secret weapon for the basis for how to actually utilize AI as a powerful competitive advantage.
• 31:58
  SPEAKER_05: This book is really the first one to combine sophisticated explanation of how AI works with a practical approach of applying AI to the problems of the business.
• 32:06
  SPEAKER_05: And it's applied to, you know, various use cases from customer experiences to business operations to product development.
• 32:13
  SPEAKER_05: It's a powerful, powerful tool in the arsenal of understanding how to deploy AI to make your business better.
• 32:18
  SPEAKER_05: The AI Powered Enterprise by Seth Early.
• 32:21
  SPEAKER_05: Harness the power of ontologies to make your business smarter, faster, more profitable.
• 32:25
  SPEAKER_05: Out on Amazon.com today.
• 32:27
  SPEAKER_05: And now back to the show.
• 32:29
  SPEAKER_02: I want to remind people that we're speaking with Michael Shrage, who is author of Recommendation Engines, uh, and he is a research fellow at the MIT Sloan School of Management's initiative on the digital economy.
• 32:42
  SPEAKER_02: So just wanted to remind folks who we're talking to today.
• 32:45
  SPEAKER_03: I didn't ask questions about the ethics of the slot, and I I'm going to hearken back to my economics background.
• 32:53
  SPEAKER_03: And there's the famous issue in economics, again, Nobel Prize winning research about the principal versus the agent, you know, or or to put it in other words, qui bono, who benefits.
• 33:06
  SPEAKER_03: If I am getting advice from somebody who I by the way, I'll really name drop here.
• 33:12
  SPEAKER_03: You remember who was on the board of the Washington Post Company when I was there?
• 33:16
  SPEAKER_03: Warren Buffett.
• 33:17
  SPEAKER_03: Oh.
• 33:17
  SPEAKER_03: So one of the perks of being a reporter at the Post was that I used to be able to talk to Mr.
• 33:22
  SPEAKER_03: Buffett.
• 33:23
  SPEAKER_03: And I had invested with him, but but it was a great experience in Stockholm.
• 33:28
  SPEAKER_03: And Mr.
• 33:29
  SPEAKER_03: Buffett was he was a good guy, great source, good guy.
• 33:33
  SPEAKER_03: And and uh, you know, he had a one, he's he's filled in his shareholder letter, which I urge everybody to read after reading my book, is is is you know just brilliantly put together.
• 33:43
  SPEAKER_03: And he had a wonderful phrase about never ask the barber if you need a haircut, you know, because there's an inherent conflict of interest.
• 33:53
  SPEAKER_03: And that's the issue that I want to deal with the question that you posed, Chris.
• 33:57
  SPEAKER_03: Because if I feel that the recommendations that I'm getting from Amazon are to sell me shit, yeah, to sell me stuff, well, you know, F you.
• 34:09
  SPEAKER_03: You're not looking out for my best interest, you're looking out for your best interest.
• 34:14
  SPEAKER_03: So the real ethical challenge here is can organizations demonstrate that the recommendations they make, the advice they offer, truly reflect and incorporate the best interests of their customer, their client, their prospect.
• 34:33
  SPEAKER_03: And this is one of the reasons to deal with an issue that you've dealt with in past podcasts.
• 34:39
  SPEAKER_03: This is one of the reasons why when people talk about interpretability and explainability of machine learning algorithms, hell yeah, we need interpretability, explainability, and yes, yes, yes, transparency around recommendation algorithms.
• 34:56
  SPEAKER_03: I want Amazon to make absolutely clear to me why they're recommending this book versus that book.
• 35:04
  SPEAKER_03: I want Spotify to make clear to me why they they are suggesting these genres of music in Spotify Discover versus that, uh others.
• 35:16
  SPEAKER_03: Same thing with LinkedIn, same thing with dare I say it, TikTok, what have you.
• 35:21
  SPEAKER_03: So interpretability, explainability, transparency matters a lot.
• 35:27
  SPEAKER_03: That, to my mind, is the best and surest way of killing off the relevance of the ethical question that you ask.
• 35:36
  SPEAKER_03: Now let's flip it to the user experience notion.
• 35:40
  SPEAKER_03: I believe recommend our systems work when they offer better choices, choices that you would not have come up with on your own.
• 35:51
  SPEAKER_03: And this is key, you can see yourself in those choices, or you can see the kind of person you want to be in those choices.
• 36:02
  SPEAKER_03: So there does have to be an affective as well as an effective uh component to recommender system design.
• 36:11
  SPEAKER_03: Because it's exactly what you said, Chris.
• 36:14
  SPEAKER_03: You want people not just to say, oh, these are great choices, you want people to feel that these are great choices, that they reflect and respect the individual receiving those choices, that advice, those suggestions, those recommendations.
• 36:31
  SPEAKER_03: That's not a subtle design problem.
• 36:34
  SPEAKER_03: That's a huge design problem.
• 36:36
  SPEAKER_03: And that's why I think recommenders are so interesting.
• 36:39
  SPEAKER_05: Well, I mean, we get into scenarios all the time too where people forget that by virtue of paying for a specific product or service, they automatically sign up for the upstream and downstream effects of you know of that product, whether it be, by the way, you have these uh devices in your house.
• 37:00
  SPEAKER_05: Yeah, they're there for home automation, but just so you know, you've also signed away the rights to say that we're gonna listen to your conversations and and reserve the right to try to recommend things for you because at the end of the day, it's a business promoting a business.
• 37:14
  SPEAKER_03: So so quick quick interruption there.
• 37:17
  SPEAKER_03: I did some advisory work for Amazon.
• 37:20
  SPEAKER_03: Yeah, I I I like them, but one of the gifts they gave me was an Alexa, which my wife refuses to use for exactly the reasons you just enumerated, Chris.
• 37:33
  SPEAKER_05: Yeah, I'm telling you, well, and I was I I was at AWS too, and we worked on the I worked on the primitives of Alexa in the Lex chatbots and stuff.
• 37:44
  SPEAKER_05: They're derivative systems now.
• 37:46
  SPEAKER_05: They don't, they they're mutually exclusive, right?
• 37:48
  SPEAKER_05: Because the difference between, let's say, a intent short form speech recognition and and chatbot intent-based framework system as opposed to a localized um system that has, let's say, DSPs and all that kind of stuff and processing globally, right?
• 38:04
  SPEAKER_05: There's there's two different things.
• 38:05
  SPEAKER_05: And plus the models are completely different.
• 38:07
  SPEAKER_05: One's made for business, the other one's made for you know consumer.
• 38:10
  SPEAKER_05: So uh, but yeah, it's you know, it's an amazing thing to think about that it scares the hell out of people in terms of what it actually is doing and what it can do.
• 38:21
  SPEAKER_03: But but Chris, you know, now I'm gonna flip this and push, make, make you the subject here, because if you really spend time with Amazon, let me let me tell you the thing that the the things I can't discuss, but this is I've discussed publicly, so I want to raise it with you.
• 38:34
  SPEAKER_03: Okay.
• 38:35
  SPEAKER_03: If I were I if I were advising them, okay, one thing I wish they would do is to import this affective aspect into Alexa.
• 38:46
  SPEAKER_03: Why I want to be able to say to an Alexa, Alexa, play me a sad song.
• 38:52
  SPEAKER_03: I want Alexa to respond, how sad.
• 38:56
  SPEAKER_03: How sad.
• 38:58
  SPEAKER_03: So that way we really do get to have these things be customized and bespoke.
• 39:03
  SPEAKER_03: Because yeah, you're spot on.
• 39:06
  SPEAKER_03: Because you know you want to build a kind of connection where I mean this is this is the issue.
• 39:12
  SPEAKER_03: If you don't trust Amazon, it'll never work.
• 39:15
  SPEAKER_03: You know, this is the key here.
• 39:18
  SPEAKER_03: You want trusted advice, and this is why we're back to being human again.
• 39:24
  SPEAKER_03: I think that that you want to design interactions and use cases, not just that offer better choices and good advice, that build trust.
• 39:37
  SPEAKER_03: It's called altogether now customer lifetime value.
• 39:42
  SPEAKER_03: Yep, yep.
• 39:43
  SPEAKER_03: Version rates boosting conversion, click-through rate is the devil.
• 39:48
  SPEAKER_03: It's the devil, it's Satan, it's Satan.
• 39:52
  SPEAKER_03: Don't yield to Satan.
• 39:55
  SPEAKER_03: Look at things with customer lifetime value, and then you can have trusted relationships that make everybody more valuable.
• 40:02
  SPEAKER_05: I'll stop thinking.
• 40:20
  SPEAKER_05: And like you said, you know, some of those metrics that are being measured on are short-sighted and therefore drive short-sighted results and outcomes that, and then they sit back and they think, well, what the hell am I doing wrong?
• 40:32
  SPEAKER_05: Well, you're not looking at the entire, like you said, the LTV of what your customers are needing to then apply it, which is a longer, longer, you know, um, insertion of all these different points of data and and information to actually pull it together and say, really, you know, you want to listen for a sad song.
• 40:52
  SPEAKER_05: And by the way, when you said that, Alexa started playing some Led Zeppelin songs that were a little bit more on the macabre side of things because I've been listening to Led Zeppelin.
• 41:00
  SPEAKER_03: Yeah, I would hope I would hope melancholy more than macabre, but yeah, melancholy, sure, sure, sure, sure.
• 41:06
  SPEAKER_05: Um, but it is Led Zeppelin, you know.
• 41:10
  SPEAKER_05: But it was, it's it's it's a point that's super important, especially with how to apply.
• 41:15
  SPEAKER_05: And I like, I think your your your topic at the very beginning of this is that it's you know, I'm trying to provide advice and personalized advice, empathetic advice, serendipic, you know, serendipity in terms of advice on how to do this.
• 41:33
  SPEAKER_05: Because to your point, your wife probably does um some interesting things with you know how she looks at purchasing, you know, different items and things, but at the same time, she doesn't want to feel like there's a monkey on her shoulder always listening to everything that she's doing to again get away from that big brother field, right?
• 41:54
  SPEAKER_05: So it's almost an invasion of privacy as opposed to us you know a recommendation of advice for these types of things for your life.
• 42:02
  SPEAKER_03: Well, I I think trust matters, but you were you were you were also going to say something, Seth?
• 42:08
  SPEAKER_02: Uh well, I was going to make a comment about um likelihood to recommend.
• 42:12
  SPEAKER_02: I think it is a valid score that has to be deconstructed, but I agree with what you're saying, Chris.
• 42:17
  SPEAKER_02: It's a timeline, it's time dependent, and it has a shorter context, whereas the uh uh lifetime value is a much longer timeline and time frame.
• 42:27
  SPEAKER_02: So I just wanted to point out that I think things like likelihood to recommend, if you do deconstruct those things, it can't just be that gross score, that big score.
• 42:36
  SPEAKER_02: It has to be deconstructed to well, what are the elements that that comprise this?
• 42:41
  SPEAKER_02: So I just wanted to say I I agree on on one hand, but also think it is important uh as a signal if you can deconstruct it and understand more about it.
• 42:50
  SPEAKER_02: But I think the more important piece here is is that that agenda, right?
• 42:55
  SPEAKER_02: Is there that hidden agenda, which in most cases there is, right?
• 42:58
  SPEAKER_02: There's not even a hidden agenda.
• 43:00
  SPEAKER_02: I want to tell you more stuff.
• 43:01
  SPEAKER_02: But when people can really appreciate you know the racity and the uh appropriateness of a recommendation, you know, that's what is going to uh positively impact that and strengthen that experience.
• 43:13
  SPEAKER_02: One of the things I wanted to ask you, Michael, is when you think about personalization, you know, there's identity personalization, there's utility personalization, but um those are but those are also really if it's utility, then it does have to have some recommendation in there.
• 43:30
  SPEAKER_02: Would you agree with that?
• 43:31
  SPEAKER_02: Like what's the line between personalization and recommendation?
• 43:34
  SPEAKER_02: I think it is, yeah.
• 43:36
  SPEAKER_03: But but you know, that but we'll go back to recommenders 101.
• 43:40
  SPEAKER_03: You know, that's the difference between that's collaborative filter filtering.
• 43:44
  SPEAKER_03: People like you get this kind of value from it.
• 43:48
  SPEAKER_03: That's the the that's a U top, but it but but it's also the content side, which is you've expressed interest in these sorts of things.
• 43:56
  SPEAKER_03: So what you're really doing is back to econ 101.
• 44:00
  SPEAKER_03: the the hybrid what's the indifference curve what's the trade-off between between these things that's that's that's the issue but i i want to go back to how you predicated this i i really believe uh it it becomes easier to be open and experiment with choices and recommendations and advice if you trust uh the person or the device or the algorithm making that recommendation and and just to to comment on on you know your book and and ontology etc that that's what's so key you have it's called data integrity it's clut the there's a lineage there that's trackable traceable accountable that what you're really talking about is is the antithesis of what we all grew up with which is garbage in garbage out okay you know that that's what I'm so intrigued by on this let I'm let me take it back that's one of the reasons why I'm so intrigued by this because the the rise of algorithmically innovative recommenders makes data makes ontology makes semantics that much more important it's that much more important absolutely absolutely the signals you're gathering the signals and harmonizing them and it has to come from some and they have to be aligned with your understanding of the user's mental model and then the models the architectural models that you're using in your technology.
• 45:38
  SPEAKER_03: Right my spin on what you just said there is I believe that we now have such a broad array broad samples of data sets and choices and offers is that the mental model has to embrace the affective as well as the cognitive and that's why you know the the Daniel Kahneman remember you know him thinking fast and slow recommenders have to be able to embrace the notion of people who reflect on recommendations and those whose reptilian brain amygdalas immediately you know respond to these reactions.
• 46:16
  SPEAKER_03: Exactly exactly and so I believe these you know it's a cliche I think recommenders are are one of the things that are turning human beings into cyborgs because increasingly uh I literally do not know anybody anywhere who drives to places they don't know without Google maps or ways.
• 46:38
  SPEAKER_03: Okay.
• 46:38
  SPEAKER_02: Right.
• 46:39
  SPEAKER_03: Yeah yeah that's a great that's a recommender it's a recommender and if you want there's stuff you can put in there that says if you are prepared to go 15 minutes out of your way you can have this view.
• 46:52
  SPEAKER_05: And and of course if you want to find an unusual restaurant etc etc so the whole notion of the of the the advisory sphere of experience the recommender sphere of experience it's it's it's multimodal it's uh it i again i find this crap fascinating absolutely fascinating and i'm astonished that so many legacy organizations underinvest in this i am astonished by the number of organizations that that that to chris's point care about nps but do such a crap job of offering advice and and now just to double down on what what what you and and and chris said they're more interested in upselling than giving you good advice when in fact the shortest way to upsell would be to give you good advice geez these people are what who are these people kidding so when do you see that then there's diminishing returns to recommendation engines because you know commenting on the cyborg you know I mean part of technology is you know at what point will people think for themselves but at the same time you know if there's always uh you know the choice is taken away and placed in front of you that you know kind of diminishes people think about what you know what I mean yeah you know one of the things that really creeped crept me out in in the book in researching the book is that all over the world and by all over the world I mean I mean uh um europe mainland China and the United States there were entrepreneurs whose philosophy of recommender design was we want to know you better than you know yourself yeah sorry that draw that crosses a line for me there I actually wrote a book of a Netflix engineer who says why are we giving people choices we should just give them the choice that we know they're they're surest to take you know so it's that's not first off that's not a recommendation that's a compliance engine that's an you know that's a do it or else engine you know I think you I think the question you're really asking is about what impact does good advice have on agency and autonomy I you know no apologies here I am a fan of technologies that promote empower and enable agency I want better choice give me good choices don't tell me the answer give me the best choices let me choose don't try to turn me into a meat puppet make me feel like somebody who is in control of my own destiny is that so wrong no I think there's I think that's and I'm glad you rephrased it that way because that's where you know an educated decision can certainly weigh out from just a uh you know blanket in the dark like I need to make a decision on something I don't know where to start so at least that gives you a starting point of options which actually can accelerate in my mind the ability to actually have autonomy and agency in a lot of ways if it's empowered in the correct way not just taking your decision away but giving you the ability to choose.
• 50:09
  SPEAKER_03: So my only pushback on that is I believe the right way I you know I do not trust in the wake of this pandemic I do not trust experts anymore.
• 50:19
  SPEAKER_03: Okay I want and I I let's make this a theme I want transparency explainability interpretability around this because that way I can see what I I want to be able to compare recommendation engines.
• 50:37
  SPEAKER_03: Yeah I want to in an ideal world and this let's here's here's our billion dollar unicorn startup let's come up with a recommender for recommenders that we look into the to the to the uh various recommended recommender systems and say if you like your advice this way these are the kinds of things you should go to so let's go meta on recommendations in the same way metacognition is thinking about thinking meta recommenders is recommendation about recommendation recommendation about advice and by the way Seth that is where I believe your book on ontology should have ended because that is what ontology enables yeah next next edition I but I agree with you that it's I look at it as personalization but I believe that we will need those recommendation engines to identify the things that are going to be most resonant with individuals and their behaviors and their needs and their mental model and their desires and their styles we're gonna have to have our own uh recommendation engines that cater to our needs and I look at that as the future of where some of these virtual assistants will will go.
• 51:46
  SPEAKER_02: So I I didn't explicitly write that um but I think it's a great it's a great concept and uh um next edition.
• 51:54
  SPEAKER_05: No I you know your books your book Seth is a is a like I you know I'm this is where we usually go with these discussions too Michael is you know how do people get started and you know Seth's book is brilliant at at how to start with understanding what data and where where to go collect it and actually how to identify it and then applying things like the application of such data like with your book and recommendations and engines works in an amazing way like in a hand to hand type of scenario.
• 52:26
  SPEAKER_03: I'd love to know too you know what what's in store for you next like what what are you working on for the next the next versions of all this so the thing that I ended the book with is really there there I have two key paths of research.
• 52:46
  SPEAKER_03: Ironically inappropriately it relates to something that you said in passing and the other is a direct extension of this work.
• 52:53
  SPEAKER_03: The one that you mentioned in passing was NPS and NPS and that promoter score is a bad metric.
• 52:58
  SPEAKER_03: In fact you know sort of the evil twin mirror image of all of the recommender system and advice data is what do we want to measure?
• 53:07
  SPEAKER_03: I'm very interested in the notion when you've got orders of magnitude more data and and algorithms that can learn what if our KPIs could learn what would it mean to have a portfolio of KPIs that could learn to optimize the portfolio okay so that's one area and you can see the other is what I refer to in the book saleswear and it ties directly into what Seth was talking about in terms of virtual assistance.
• 53:35
  SPEAKER_03: Okay well I am interested in recommendations for me no question but I'm married I have a wife um I as a rule I prefer her to be happier rather than unhappy so I get recommendations I I think so I think so how should I weight how should I weight her mental model her affect her cognition in the recommendations that I get on where we should travel where we should eat you know how much time to spend with the relatives you know these are not rhetorical questions Netflix true story Netflix has spent tens of millions trying to refine recommenders for families as opposed to individual accounts right this is huge for them it's huge for them again not hypothetical so I'm interested in the notion my argument is human beings are not self you're not yourself you are the sigma and sum of of multiple selves what does it mean to make a recommendation for your curious self versus your cautious self what does that mean by the way what's the overlap you know Jacques Art similarity what is the overlap between advice that you would give your adventuresome self versus the advice you'd give to your typical self yeah a high overlap or a low overlap so I'll cut to the change recommenders are about the future of introspection and understanding yourself and that really excites me even if it sounds narcissistic sorry no it's not at all I mean it's it's I think you nailed it down onto that a recommendation engine is to the individual and if you can figure out how to scale to the individual then you've got something there that can actually take those into consideration.
• 55:37
  SPEAKER_05: If you just do it on a mass you know then you'll miss you know a ton of it but it's it's I totally agree.
• 55:44
  SPEAKER_02: Hey listen we're at the top of the hour this has gone so fast this has been such a pleasure Michael I really always enjoy you know the conversations and interactions we've had and uh we definitely have to continue this conversation but this has been a blast.
• 56:00
  SPEAKER_03: I I am so grateful for this incredibly wide ranging and fun conversation and and I I do I'm you've given me a lot to think about in this regard and and I'm I'm I I enjoyed it and I look forward to whatever feedback materializes.
• 56:16
  SPEAKER_03: Thank you so much.
• 56:17
  SPEAKER_05: Before we close I'd like to remind everybody about our sponsors Simpler Media, Marketing AI Institute and Early Information Science.
• 56:24
  SPEAKER_05: For more information check the show notes below for links to their respective sites.
• 56:32
  SPEAKER_00: Thanks for listening to the EarlyAI podcast with your host early and Chris Featherstone we hope you enjoyed the program and took away some ideas that you can start to implement in your organization right now.
• 56:44
  SPEAKER_00: If you still need help check out www.early.com for information about early information tonight.
• 56:52
  SPEAKER_00: And of course check the show notes for links to any resources we mentioned today.
• 56:57
  SPEAKER_00: That's all for now see you next time