Applied Anthropology Reveals Surprise Insights AI Cannot Replicate
TL;DR
- True insight emerges from the gap between perceived reality and actual reality, often revealed through unexpected, painful moments of struggle that challenge deeply held assumptions.
- Applied anthropology uncovers non-obvious human behaviors and motivations by immersing researchers in natural environments, providing a "pre-hypothesis" understanding distinct from psychology's focus on individual minds.
- AI's current limitations lie in its lack of embodied experience, sensory input, and nuanced social understanding, preventing it from grasping the full spectrum of human emotion and context.
- Companies can leverage AI to accelerate experimentation by rapidly simulating customer behaviors and market responses, thereby increasing the pace of innovation beyond traditional research methods.
- The "surprise" of genuine insight is difficult for AI to replicate because it requires a subjective, embodied experience of encountering the unexpected, a process often fueled by struggle and doubt.
- AI-generated ideas are often objectively superior but are less likely to be selected by humans who overvalue their own contributions due to familiarity and the "IKEA effect" of personal investment.
- Understanding human culture requires observing embodied actions and social dynamics, not just analyzing language, as seen in how children's play reveals complex social hierarchies and self-discovery.
Deep Dive
True insight into human behavior and innovation stems not from predictable data, but from unexpected gaps between assumptions and reality, often revealed through embodied experience and struggle. While AI can accelerate pattern recognition, it currently lacks the human capacity for surprise, embodied understanding, and the nuanced interpretation of social context, limiting its ability to generate breakthrough insights.
Applied anthropology, as practiced by Mikkel B. Rasmussen, focuses on pre-hypothesis science: immersing researchers in real-world environments to observe and understand human culture and social dynamics. This "fieldwork" involves participatory observation, where anthropologists engage directly with subjects--children, doctors, drivers--to grasp the meaning of their activities from their perspective, rather than imposing pre-existing theories. This approach differs from psychology by studying humans as social beings within groups, rather than focusing on individual minds. The core of this practice is uncovering "surprise"--moments where observed reality contradicts deeply held assumptions, leading to genuine insight. These moments are often preceded by pain, doubt, and struggle, indicating a deeper, more complex understanding is emerging.
The implications for AI are significant. While AI excels at processing vast amounts of linguistic data and identifying statistical patterns, it struggles to grasp the non-linguistic dimensions of human experience: the sensory, the emotional, and the embodied. AI lacks a body and therefore cannot understand how things feel, smell, or look in the way humans do. Furthermore, human sociality--the deep interconnectedness and cultural conditioning we inherit--is not merely language-based; it is embedded in social pressure and learned behaviors. Consequently, AI's current understanding is limited to descriptive knowledge, not experiential knowing. For example, while an AI can process descriptions of hunger, it cannot truly "know" hunger without the embodied experience. This distinction highlights a critical gap: AI can analyze what humans say, but it cannot fully grasp what they do or feel without embodied context.
This gap creates opportunities for synergy. Corporations hire applied anthropologists when they lack hypotheses about complex problems, such as LEGO's near-bankruptcy 20 years ago. By studying why children play, rather than just how, LEGO discovered that play is about depth, mastery, and self-discovery, not instant satisfaction. This insight led to a radical product roadmap shift, cutting 70% of products and focusing on deeper engagement. Similarly, AI can be trained to assist in understanding human behavior by analyzing vast datasets, including video and audio. However, the critical "surprise moments" that drive breakthrough innovation still require human interpretation, driven by embodied experience and the struggle to reconcile conflicting data. The ambition of "anthropology without anthropologists" aims to use AI for unbiased observation and pattern recognition, but the ultimate synthesis and the generation of novel insights remain a human prerogative, particularly the ability to connect disparate observations into a profound, embodied epiphany.
Action Items
- Audit AI assumptions: For 3-5 core AI initiatives, document current assumptions about human behavior and compare them to observed reality.
- Create "anthropology without anthropologists" framework: Define 5 key AI capabilities for unbiased human behavior analysis using video and audio data.
- Design AI-driven interview protocol: Develop prompts for AI to elicit nuanced insights on human experience, focusing on sensory and social dimensions.
- Measure AI-generated insights vs. human-generated insights: For 3-5 specific use cases, quantify the difference in surprise and depth between AI and human analysis.
- Develop embodied AI training data strategy: Identify 3-5 critical human sensory and social experiences to prioritize for synthetic data generation.
Key Quotes
"Applied anthropology which means studying human culture and studying human beings and particularly their social world so what is play what is illness what is traffic what is ai and how do we construct the world around that and it's used to understand the meaning of things and particularly in in corporate world corporations use it when they have no hypothesis about the problem they're solving particularly so it's what's called pre hypothesis science"
Mikkel Rasmussen explains that applied anthropology focuses on understanding human culture and social worlds, rather than individual minds. He highlights its use in corporate settings when companies lack initial hypotheses, likening it to pre-hypothesis science where observation precedes theory.
"language is only one dimension of human nature there is also the body like how does thing feel how does the sensory system so you know how does things smell so try to explain how something smells is very very difficult to do with language or how does something look how does you know early morning in october in copenhagen look it's you can do it but you need almost to be like a poet to describe because it's not just you know words it's also emotion and and and what you see with your eyes and then there's the whole thing around how things feel with your hands so the sensory and there's a whole lot of science on this that they are basically it says that you know we don't think with our brains alone we think with our bodies and that's probably something that is not yet fully understood by ai or where ai has a little bit of a weakness because it doesn't have a body yet"
Rasmussen points out that language, while important, is only one aspect of human experience. He emphasizes the role of the body and sensory input--how things feel, smell, and look--in human understanding, suggesting that AI currently struggles with these embodied experiences because it lacks a physical form.
"all great insights come from a gap between how you think the world is and what it really is what reality is so in all companies there are these assumptions that you built your business around for example instant traction in the case of lego and then there's reality which says the opposite are often and it's that gap"
Rasmussen defines insight as arising from the discrepancy between a company's assumptions about the world and the actual reality. He uses Lego's initial assumption of "instant traction" for toys as an example, contrasting it with the reality of children's deeper engagement with play.
"surprise is always there's a question of relative to what yeah and you and just even as you're talking about surprise i realize it's relative to my own expectations and understanding going back to your definition of insight right yeah there's the gap between what i think the world is and and what it actually is but the important thing there is i and the reason perhaps that ai can't deliver the surprise i'm just riffing right now but it can't deliver the same surprises it has a poor approximation of i"
The speaker discusses how surprise, a key element of insight, is relative to one's own expectations and understanding. They posit that AI may struggle to deliver genuine surprise because it lacks a true "self" or a personal framework of expectations against which to measure deviations from reality.
"i think there's a general tendency towards what's called anthropomorphizing i think it's called when you assume that the machine you're building is human so you give it a name and you call it it and him and her and so on i think that's a mindset that i probably will disappear over time when ai becomes more a natural part of our world and then i don't think we'll even discuss you know what's human what's machine because it will be very obvious a little bit like i think some people on the podcast have described it as electricity yeah and we don't we don't talk about it it makes it light up in my room right it's just like i turn on the lights and it's there"
Rasmussen suggests that the tendency to anthropomorphize AI, by giving it human names and pronouns, is a mindset that will likely fade as AI becomes more integrated into daily life. He compares this future integration to electricity, which is now a taken-for-granted utility rather than a subject of constant discussion.
"i think there's one thing that's important to say which is part of doing what we do what i do is also an embodied process it really is so when i work with a ceo or leadership it's important that they see the field or the people that we're studying themselves and that they get a embodied sense of surprise it's not just intellectual that is that also"
Rasmussen emphasizes that his work, particularly when collaborating with CEOs and leadership, involves an embodied process. He believes it is crucial for leaders to personally witness the subjects of study to experience an "embodied sense of surprise," which goes beyond mere intellectual understanding.
Resources
External Resources
Books
- "Sensemaking" by Christian Aspberg - Mentioned in relation to big data and its pre-AI articulation of concepts AI struggles with.
- "A Technique for Producing Ideas" by James Webb Young - Mentioned as a book on creativity that discusses the ingredient of "hopelessness" in the idea generation process.
People
- Mikkel B. Rasmussen - Founder of Human Activity Laboratory, applied anthropologist, and guest on the podcast discussing anthropology and AI.
- Christian Aspberg - Colleague of Mikkel B. Rasmussen, co-founder of "Anthropology without Anthropologists," and author of "Sensemaking."
- Jan LeCun - Mentioned as a figure whose criticism of large language models suggests they do not understand sensory experiences like smell or taste.
- Darwin - Referenced as an example of pre-hypothesis science, akin to applied anthropology.
Organizations & Institutions
- Human Activity Laboratory - Mikkel B. Rasmussen's organization focused on understanding anthropology and AI.
- Meta - Mentioned as the organization Christian Keller joined.
- Google - Mentioned for their work on Pixel phone's contextual video features.
- Lego - Discussed as a company that used applied anthropology to understand why children play, leading to a significant shift in their product strategy.
- IKEA - Mentioned in relation to the "IKEA effect," where personal effort increases perceived value.
Other Resources
- Anthropology without Anthropologists - A project by Mikkel B. Rasmussen and Christian Aspberg focused on using AI for insights into human activity.
- AGI (Artificial General Intelligence) - Discussed in the context of embodied learning and sensory experience.
- Large Language Models (LLMs) - Discussed as a type of AI that understands the world through language but may lack sensory and embodied understanding.
- Contextual Video - A feature on Google's Pixel phone that can recognize objects and understand context.
- Embodied Learning - A concept related to AI understanding through physical interaction and sensory experience.
- Synthetic Data - Discussed as a tool for solving smaller problems, A/B testing, and potentially accelerating experimentation and innovation.
- Thick Data/Thick Description - The detailed, multi-faceted understanding of human behavior that AI currently struggles to replicate.
- AI Agents - Mentioned as an interesting area for studying how children interact with AI.
- Instant Traction - A concept Lego previously held about satisfying toys quickly, contrasted with the deeper nature of play.
- Participatory Observation - A scientific term for fieldwork where one participates in and studies a social activity.
- Fieldwork - The practice of studying a particular group of people by being with and observing them.
- Pattern Recognition - An area where AI is used to analyze data, though it cannot yet create the "aha" moment of surprise.
- Multimodal Data - Data from various sources like video, audio, text, and sales reports, which can be synthesized for understanding.