{"title":"丹尼尔·卡尼曼对我思想的影响","authors":"Gary Marcus","doi":"10.1002/aaai.70108","DOIUrl":null,"url":null,"abstract":"<p>It would not be an exaggeration to say that Danny Kahneman has influenced my thinking throughout my adult life and even earlier; I first learned about one strand of his work at a summer class on cognitive psychology when I was fifteen. Just a few years later, I became completely immersed in it. When I got to know him personally, in his later years, it was a huge thrill. We would often have lunch at the Pearl Oyster Bar in Greenwich Village, both of us invariably ordering their amazing lobster rolls. The restaurant is gone and I miss those lobster rolls, but I miss Danny much more.</p><p>The last time that I saw Danny was just a couple of months before he died, at a dinner, and he had switched name tags so that we could sit together. I will always cherish that memory. It was also a real thrill to have him participate in the 20-or so person “AI debate” that I co-hosted during the pandemic; he participated reluctantly, saying he wasn't sure he had anything to contribute, and characteristically was lucid and a highlight of the night.</p><p>Most of our conversations were about AI, which I had gravitated towards and which he was deeply curious about. But they were built on and grounded in a shared interest in cognitive psychology.</p><p>And it is really through cognitive psychology rather than AI that I first became aware of his work, and how his work first began to influence me. I had heard brief mention of his work on human irrationality in that summer class. By the time I decided to go to Hampshire College a few months later, I already knew that I wanted to study cognitive science, drinking as deeply as I could from fields like philosophy, psychology, linguistics, and computer science.</p><p>In a summer project at Hampshire, supervised by Neil Stillings, I studied reasoning from a cross-cultural perspective; though most of Danny's work was on Western subjects, a large fraction of what I read was influenced by Danny. A 1982 book that he co-edited with Amos Tversky and Paul Slovic, <i>Judgment Under Uncertainty: Heuristics and Biases</i> was practically my bible (Tversky & Kahneman, <span>1974</span>). My undergraduate thesis (known at Hampshire as a “Division III”) was an experimental study about one facet of rationality: how people formed and revised their beliefs relative to evidence — very much a Danny kind of question. (In a nutshell I found that people quickly lost track of how they formed their beliefs, and would sometimes develop rationalizations for those beliefs, retaining those rationalized beliefs even if the original evidence for them was discredited.)</p><p>In graduate school, I left reasoning behind and went to work with Steven Pinker, on an entirely different topic: language acquisition. But the things I found there in some ways paralleled the “System I” (fast, reflexive, automatic) and “System II” (deliberative reasoning) distinction that Danny later made famous in <i>Thinking, Fast and Slow</i> (Kahneman, <span>2011</span>). My first major project as a graduate student, which became the core of my dissertation, showed that children used two distinct mechanisms in acquiring the past tense of English, an associative memory system that was statistical in nature for irregular verbs (e.g., <i>sing-sang, go-went</i>), and a rule-following system for regular verbs (<i>talk-talked</i>, <i>walk-walked</i>, etc.). There were clearly parallels with what Danny described.</p><p>Danny's work influenced mine in many ways, throughout my career. Another point of intersection, less focused on human (ir)rationality, came in my 2001 book The Algebraic Mind (Marcus, <span>2003</span>), where I laid out the arguments for neurosymbolic AI, from a cognitive science perspective—and first warned about hallucinations, which of course haunt large language models to this day. I showed how hallucinations emerged in an experiment with a neural network model that was popular then, a predecessor to today's models. I argued that that network lacked an adequate distinction between individuals and kinds, and showed how it overgeneralized information about individuals to other people that shared properties. Instead, I argued that people used something like Danny's notion (in work with Anne Treisman) of object files. You could think of these as being something like index cards, holding knowledge about specific entities. My argument then was that until we developed alternative neural networks with something similar, the kind of overgeneralization errors that I observed (that we now know as hallucinations) would likely be inevitable. A quarter century later, I still see it that way; modern models are better in many ways, but still lack proper object files.</p><p>My 2008 book <i>Kluge: The Haphazard Evolution of the Human Mind</i> (which Danny read and enjoyed, before we became friends, sending me an unsolicited email that made my day!) was about human irrationality, and how it evolved (Marcus, <span>2009</span>). It was the first time that I wrote explicitly about the System I/System II distinction (which I called reflexive vs deliberative), and Danny's influence can be seen throughout the book; in essence I was trying to explain <i>why</i> the human mind was littered with the kinds of biases that Danny, and others influenced by him, had documented for decades. (In a nutshell, the argument of the book was that the human mind evolved not as a perfect solution but as an aggregation on top of prior systems that were better suited to other purposes, such as tracking overall statistical trends.)</p><p>My 2023 TED Talk used Danny's famous System I and System II distinction to frame the need for neurosymbolic AI (Marcus, 2023). I likened neural networks to his System I, classical AI to System II, and said that it was obvious that AI would need both. (Recent systems like Claude Code and OpenAI's codex do in fact use both; for a long time, the field tried to get by with one or the other.)</p><p>The thing is, I am sure thousands of other scientists could tell similar stories, about how deeply Danny's work influenced them. Danny influenced cognitive psychology, economics, anthropology, politics, AI and so many other fields; he traversed them all fluently, and of course won a Nobel in Economics even though much of his academic career was spent in psychology departments. (There is no Nobel for psychology, but he surely would have won one had there been.) It is a true testament to his reach that The Economist listed him as the seventh most influential economist in the world when Danny wasn't even trained as an economist.</p><p>The last time I saw Danny was in November 2023; he was 89, passionate, curious, and completely lucid, one of the sharpest minds I have ever met.</p><p>I miss him dearly.</p><p></p><p><i>The author with Daniel Kahneman and Hans Zimmer, November 2023, New York City</i></p><p>The author declares no conflicts of interest.</p>","PeriodicalId":7854,"journal":{"name":"Ai Magazine","volume":"47 4","pages":""},"PeriodicalIF":3.5000,"publicationDate":"2026-10-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1002/aaai.70108","citationCount":"0","resultStr":"{\"title\":\"Daniel Kahneman's Influence on My Thought\",\"authors\":\"Gary Marcus\",\"doi\":\"10.1002/aaai.70108\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p>It would not be an exaggeration to say that Danny Kahneman has influenced my thinking throughout my adult life and even earlier; I first learned about one strand of his work at a summer class on cognitive psychology when I was fifteen. Just a few years later, I became completely immersed in it. When I got to know him personally, in his later years, it was a huge thrill. We would often have lunch at the Pearl Oyster Bar in Greenwich Village, both of us invariably ordering their amazing lobster rolls. The restaurant is gone and I miss those lobster rolls, but I miss Danny much more.</p><p>The last time that I saw Danny was just a couple of months before he died, at a dinner, and he had switched name tags so that we could sit together. I will always cherish that memory. It was also a real thrill to have him participate in the 20-or so person “AI debate” that I co-hosted during the pandemic; he participated reluctantly, saying he wasn't sure he had anything to contribute, and characteristically was lucid and a highlight of the night.</p><p>Most of our conversations were about AI, which I had gravitated towards and which he was deeply curious about. But they were built on and grounded in a shared interest in cognitive psychology.</p><p>And it is really through cognitive psychology rather than AI that I first became aware of his work, and how his work first began to influence me. I had heard brief mention of his work on human irrationality in that summer class. By the time I decided to go to Hampshire College a few months later, I already knew that I wanted to study cognitive science, drinking as deeply as I could from fields like philosophy, psychology, linguistics, and computer science.</p><p>In a summer project at Hampshire, supervised by Neil Stillings, I studied reasoning from a cross-cultural perspective; though most of Danny's work was on Western subjects, a large fraction of what I read was influenced by Danny. A 1982 book that he co-edited with Amos Tversky and Paul Slovic, <i>Judgment Under Uncertainty: Heuristics and Biases</i> was practically my bible (Tversky & Kahneman, <span>1974</span>). My undergraduate thesis (known at Hampshire as a “Division III”) was an experimental study about one facet of rationality: how people formed and revised their beliefs relative to evidence — very much a Danny kind of question. (In a nutshell I found that people quickly lost track of how they formed their beliefs, and would sometimes develop rationalizations for those beliefs, retaining those rationalized beliefs even if the original evidence for them was discredited.)</p><p>In graduate school, I left reasoning behind and went to work with Steven Pinker, on an entirely different topic: language acquisition. But the things I found there in some ways paralleled the “System I” (fast, reflexive, automatic) and “System II” (deliberative reasoning) distinction that Danny later made famous in <i>Thinking, Fast and Slow</i> (Kahneman, <span>2011</span>). My first major project as a graduate student, which became the core of my dissertation, showed that children used two distinct mechanisms in acquiring the past tense of English, an associative memory system that was statistical in nature for irregular verbs (e.g., <i>sing-sang, go-went</i>), and a rule-following system for regular verbs (<i>talk-talked</i>, <i>walk-walked</i>, etc.). There were clearly parallels with what Danny described.</p><p>Danny's work influenced mine in many ways, throughout my career. Another point of intersection, less focused on human (ir)rationality, came in my 2001 book The Algebraic Mind (Marcus, <span>2003</span>), where I laid out the arguments for neurosymbolic AI, from a cognitive science perspective—and first warned about hallucinations, which of course haunt large language models to this day. I showed how hallucinations emerged in an experiment with a neural network model that was popular then, a predecessor to today's models. I argued that that network lacked an adequate distinction between individuals and kinds, and showed how it overgeneralized information about individuals to other people that shared properties. Instead, I argued that people used something like Danny's notion (in work with Anne Treisman) of object files. You could think of these as being something like index cards, holding knowledge about specific entities. My argument then was that until we developed alternative neural networks with something similar, the kind of overgeneralization errors that I observed (that we now know as hallucinations) would likely be inevitable. A quarter century later, I still see it that way; modern models are better in many ways, but still lack proper object files.</p><p>My 2008 book <i>Kluge: The Haphazard Evolution of the Human Mind</i> (which Danny read and enjoyed, before we became friends, sending me an unsolicited email that made my day!) was about human irrationality, and how it evolved (Marcus, <span>2009</span>). It was the first time that I wrote explicitly about the System I/System II distinction (which I called reflexive vs deliberative), and Danny's influence can be seen throughout the book; in essence I was trying to explain <i>why</i> the human mind was littered with the kinds of biases that Danny, and others influenced by him, had documented for decades. (In a nutshell, the argument of the book was that the human mind evolved not as a perfect solution but as an aggregation on top of prior systems that were better suited to other purposes, such as tracking overall statistical trends.)</p><p>My 2023 TED Talk used Danny's famous System I and System II distinction to frame the need for neurosymbolic AI (Marcus, 2023). I likened neural networks to his System I, classical AI to System II, and said that it was obvious that AI would need both. (Recent systems like Claude Code and OpenAI's codex do in fact use both; for a long time, the field tried to get by with one or the other.)</p><p>The thing is, I am sure thousands of other scientists could tell similar stories, about how deeply Danny's work influenced them. Danny influenced cognitive psychology, economics, anthropology, politics, AI and so many other fields; he traversed them all fluently, and of course won a Nobel in Economics even though much of his academic career was spent in psychology departments. (There is no Nobel for psychology, but he surely would have won one had there been.) It is a true testament to his reach that The Economist listed him as the seventh most influential economist in the world when Danny wasn't even trained as an economist.</p><p>The last time I saw Danny was in November 2023; he was 89, passionate, curious, and completely lucid, one of the sharpest minds I have ever met.</p><p>I miss him dearly.</p><p></p><p><i>The author with Daniel Kahneman and Hans Zimmer, November 2023, New York City</i></p><p>The author declares no conflicts of interest.</p>\",\"PeriodicalId\":7854,\"journal\":{\"name\":\"Ai Magazine\",\"volume\":\"47 4\",\"pages\":\"\"},\"PeriodicalIF\":3.5000,\"publicationDate\":\"2026-10-26\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://onlinelibrary.wiley.com/doi/epdf/10.1002/aaai.70108\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Ai Magazine\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://onlinelibrary.wiley.com/doi/10.1002/aaai.70108\",\"RegionNum\":4,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Ai Magazine","FirstCategoryId":"94","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1002/aaai.70108","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
It would not be an exaggeration to say that Danny Kahneman has influenced my thinking throughout my adult life and even earlier; I first learned about one strand of his work at a summer class on cognitive psychology when I was fifteen. Just a few years later, I became completely immersed in it. When I got to know him personally, in his later years, it was a huge thrill. We would often have lunch at the Pearl Oyster Bar in Greenwich Village, both of us invariably ordering their amazing lobster rolls. The restaurant is gone and I miss those lobster rolls, but I miss Danny much more.
The last time that I saw Danny was just a couple of months before he died, at a dinner, and he had switched name tags so that we could sit together. I will always cherish that memory. It was also a real thrill to have him participate in the 20-or so person “AI debate” that I co-hosted during the pandemic; he participated reluctantly, saying he wasn't sure he had anything to contribute, and characteristically was lucid and a highlight of the night.
Most of our conversations were about AI, which I had gravitated towards and which he was deeply curious about. But they were built on and grounded in a shared interest in cognitive psychology.
And it is really through cognitive psychology rather than AI that I first became aware of his work, and how his work first began to influence me. I had heard brief mention of his work on human irrationality in that summer class. By the time I decided to go to Hampshire College a few months later, I already knew that I wanted to study cognitive science, drinking as deeply as I could from fields like philosophy, psychology, linguistics, and computer science.
In a summer project at Hampshire, supervised by Neil Stillings, I studied reasoning from a cross-cultural perspective; though most of Danny's work was on Western subjects, a large fraction of what I read was influenced by Danny. A 1982 book that he co-edited with Amos Tversky and Paul Slovic, Judgment Under Uncertainty: Heuristics and Biases was practically my bible (Tversky & Kahneman, 1974). My undergraduate thesis (known at Hampshire as a “Division III”) was an experimental study about one facet of rationality: how people formed and revised their beliefs relative to evidence — very much a Danny kind of question. (In a nutshell I found that people quickly lost track of how they formed their beliefs, and would sometimes develop rationalizations for those beliefs, retaining those rationalized beliefs even if the original evidence for them was discredited.)
In graduate school, I left reasoning behind and went to work with Steven Pinker, on an entirely different topic: language acquisition. But the things I found there in some ways paralleled the “System I” (fast, reflexive, automatic) and “System II” (deliberative reasoning) distinction that Danny later made famous in Thinking, Fast and Slow (Kahneman, 2011). My first major project as a graduate student, which became the core of my dissertation, showed that children used two distinct mechanisms in acquiring the past tense of English, an associative memory system that was statistical in nature for irregular verbs (e.g., sing-sang, go-went), and a rule-following system for regular verbs (talk-talked, walk-walked, etc.). There were clearly parallels with what Danny described.
Danny's work influenced mine in many ways, throughout my career. Another point of intersection, less focused on human (ir)rationality, came in my 2001 book The Algebraic Mind (Marcus, 2003), where I laid out the arguments for neurosymbolic AI, from a cognitive science perspective—and first warned about hallucinations, which of course haunt large language models to this day. I showed how hallucinations emerged in an experiment with a neural network model that was popular then, a predecessor to today's models. I argued that that network lacked an adequate distinction between individuals and kinds, and showed how it overgeneralized information about individuals to other people that shared properties. Instead, I argued that people used something like Danny's notion (in work with Anne Treisman) of object files. You could think of these as being something like index cards, holding knowledge about specific entities. My argument then was that until we developed alternative neural networks with something similar, the kind of overgeneralization errors that I observed (that we now know as hallucinations) would likely be inevitable. A quarter century later, I still see it that way; modern models are better in many ways, but still lack proper object files.
My 2008 book Kluge: The Haphazard Evolution of the Human Mind (which Danny read and enjoyed, before we became friends, sending me an unsolicited email that made my day!) was about human irrationality, and how it evolved (Marcus, 2009). It was the first time that I wrote explicitly about the System I/System II distinction (which I called reflexive vs deliberative), and Danny's influence can be seen throughout the book; in essence I was trying to explain why the human mind was littered with the kinds of biases that Danny, and others influenced by him, had documented for decades. (In a nutshell, the argument of the book was that the human mind evolved not as a perfect solution but as an aggregation on top of prior systems that were better suited to other purposes, such as tracking overall statistical trends.)
My 2023 TED Talk used Danny's famous System I and System II distinction to frame the need for neurosymbolic AI (Marcus, 2023). I likened neural networks to his System I, classical AI to System II, and said that it was obvious that AI would need both. (Recent systems like Claude Code and OpenAI's codex do in fact use both; for a long time, the field tried to get by with one or the other.)
The thing is, I am sure thousands of other scientists could tell similar stories, about how deeply Danny's work influenced them. Danny influenced cognitive psychology, economics, anthropology, politics, AI and so many other fields; he traversed them all fluently, and of course won a Nobel in Economics even though much of his academic career was spent in psychology departments. (There is no Nobel for psychology, but he surely would have won one had there been.) It is a true testament to his reach that The Economist listed him as the seventh most influential economist in the world when Danny wasn't even trained as an economist.
The last time I saw Danny was in November 2023; he was 89, passionate, curious, and completely lucid, one of the sharpest minds I have ever met.
I miss him dearly.
The author with Daniel Kahneman and Hans Zimmer, November 2023, New York City
期刊介绍:
AI Magazine publishes original articles that are reasonably self-contained and aimed at a broad spectrum of the AI community. Technical content should be kept to a minimum. In general, the magazine does not publish articles that have been published elsewhere in whole or in part. The magazine welcomes the contribution of articles on the theory and practice of AI as well as general survey articles, tutorial articles on timely topics, conference or symposia or workshop reports, and timely columns on topics of interest to AI scientists.