Tools for Thought as Cultural Practices, not Computational Objects

Here’s some light context about me:

• Designer, mediocre dev, anthropologist
• Design engineer at a research and design agency called Normally. We do early-stage design and prototyping work. Mostly exploratory AI products and capabilities for larger companies.
• Enthusiastic about visual programming, EUP, metaphors.

Who here has heard the phrase “tools for thought”?

Considering this is a lecture series on tools for augmented thinking and you are mostly MIT people, I’m going to be talking to you as potential creators of tools for thought, even if you don’t know much about this specific phrase yet. I’ll get into it’s meaning and history in a minute.

I’m also going to assume a little more computer science background than most other places, but I’ll give you all links and citations to everything I mention at the end so you can look it up later.

Tools for thought is a few things

It’s first and foremost a philosophical dream with a long history. About what happens when you give humans really well designed tools that expand the kind of thoughts they’re able to think.

Second, in recent years, it’s become a gathering point.It unifies a number of otherwise disparate fields.

And most interestingly, over the last 5 years, it’s become a category of software in the spreadsheets of VC firms and a thing that they’re willing to throw cash at

The recent hype around this phrase first starting building in 2019.

Around the time that two fairly well-known researchers - Andy Matuschak and Michael Neilson - wrote this very long, comprehensive essay on the history of tools for thought, which gave it lot of renewed interest.

In it they gave readers a sweep through a lot of major thinkers and events in computing history like Douglas Engelbart’s work and Alan Kay, and made a plea for more people to start exploring this space of computing systems that augment the human intellect.

Around the same time (2019) a new note taking app called Roam Research stared to gain traction and aligned itself with this phrase “tools for thought”.

It looked pretty ugly, and had terrible performance, but it did some things that at the time seemed novel:

• Bidirectional links or backlinks
• Used an outliner format where each piece of content is a single node that can be linked to any other node and embedded anywhere else in the graph.

They also has this crazy graph view that showed you connections between all your notes And the whole thing was multiplayer.

The founder were experimenting with what seemed like novel ways of interacting with hypertext (but you might recognise some of these are quite old). Had precedence in the 80s and 90s, particularly Ted Nelson’s work on Xanadu.

Roam was trying to resurface these. And they had a compelling goal: giving people better tools for developing ideas collaboratively and synthesising them. Trying to enable bottom-up thinking and emergence.

Roam kicked off a bit of a note-taking hype cycle. Suddenly we got a bunch of new note-taking startups that all proclaimed to be new tools for thought

This is Muse - more canvas and drawing based

Anytype - more focused on structured data

Kosmik

Reflect promising you can “think better” with this. Even if they don’t use the phrase “tools for thought” directly, creators of these tools understood they were now supposed to be in the business of making people think better

Clover

Logseq promising increasing understanding

There’s a bunch more of these - all released within the last 5 years, with the exception of Notion.

The phrase now feels synonymous with these experimental note-taking apps.

All promising their features would lead to better thinking; better quality thoughts, better memory, and more emergent connections between topics

We also got an endless pile of influencers making YouTube videos teaching you how to use these apps and setup elaborate organisational systems

This surge in interested let to more funding. Betaworks hosted an accelerator for a dozen tools for thought companies in New York

Essentially, Tools for thought became an idea machine Idea machines are a concept Nadia Asparouhova coined in an essay in 2022.

This means an ideology with backed by raw, tangible resources. Meaning people, companies, software, and money. She used examples like effective altruism or Web3, but frankly I think the Catholic Church also counts.

They are ideas with power behind them. Something we can name and point at and gather resources around.

And I’m glad tools for thought has become a financially-backed, powerful movement, because it’s a worthwhile cause.

BUT, I think the current understanding of it has some limitations and needs some redirection.

So here’s the big question this talk is asking. What if we consider…

I’m coming at this from an anthropology and design perspective And I’m very much part of the community of people using and building this kind of software. I was an early user of Roam. I know a lot of the people building these TFT companies, so I feel obliged to critically interrogate it a bit.

The term and the community around it has grown significant enough that we’re all walking around with a lot of assumptions about the meaning and mission of this phrase, and what kind of work it requires.

But thinking about different ways we might interpret this term could lead us down very different paths about what we choose to work on and choose to build

So one of the first critical questions we should ask is what does TFT have to do with computers? The phrase itself implies nothing about computation or digital objects

It’s simply tools that helps humans think new kinds of thoughts

If we take that definition at face value, and try to come up with examples of TFT, we’d have to conclude most of them precede computing by thousands of years

All of these are pre-computational

Hindu-arabic numerals made accounting and record keeping possible. Calculations could be stored, shared, and referenced Enabled all bureaucracy since.

We’ve used poetry and song to extend memory and pass stories between generations before written language

Bards used rhetorical techniques and metric rhythm to make the stories easy to recite.

It’s how pieces like the Epic of Gilgamesh and the Mahabrata survived long enough to eventually be written down

Maps enabled humans to mentally navigate space differently - gave us allocentric views vs. egocentric.

We’re able to construct mental models of much larger spaces that we can’t directly see

Cartesian coordinates; famously developed by Rene Descartes after watching a fly circle around his room

Made us able to describe points in space with specific numbers. Allows us to model objects that don’t exist with accuracy. Allows us to precisely map and refer to points on the earth. Led to GPS.

Zettelkasten (slip box) A technique from around the 18th century where you write atomic notes on individual index cards and keep them all in a large box or “slip box”. You assign each card an ID, and use those to link cards together to create deeply interconnected ideas.

Pre-cursor to deeply linked content we have with hyperlinking.

But this is primarily a list of cultural practices and systems; they are techniques – a way of doing things - that enable us to relate to information and representation in new ways. Thus enabling new kinds of thoughts.

They aren’t technological objects like hammers or compasses. They’re practices that people developed and passed down to one another through culture.

Most of the things we’re all enamoured with in the current software moment around tools for thought are actually the practices we do inside the new tools, rather than the tools themselves

Most of the YouTube thought leaders turn out to be teaching specific practices – like building a Zettelkasten – rather than teaching us specific software.

And it’s unclear how much enabling the software is doing vs. the cultural practice

The same goes for spaced repetition - there are guides on how to do it in every platform imaginable.

But we can’t quite tell if the app or the practice is what’s enabling people to improve their memory

So we’re in an interesting place; all this software is being called tools for thought And most of the community is focused on building more software

But it seems like these cultural practices are the real tools for thought

What we’re exploring with here is the relationship between the computer as a medium and the cultural practices we do within the medium. And which one of those we should be focusing on…

So we’re still on this question. What does any of this have to do with computers?

To understand why we’re confused it’s worth going back to the origins of the phrase and the people who proposed we should start paying attention to the idea.

TFT as a concept is rooted in computer science and HCI.

The first person to use the phrase was Kenneth Iverson talked about tools for thought throughout the 1950s-1960s wrote “Notation as a Tool for Thought, published by the ACM in 1980

Talking about mathematical notation as the best known and best developed example of a tool for thought.

Around the same time period of the 1950’s and 60’s Douglas Englebart wrote “Augmenting the Human Intellect”

His thesis was computers should help us approach complex situations

Alan Kay picked up a lot of this Philosophy and applied it while developing the foundations of most of our current computing environments at Xerox PARC Wrote a paper called “User Interface: A Personal View” in 1989 where he talked about his vision of the potential of computational mediums

Summed up in this quote

There’s a book called ‘tools for thought’ Howard Rheingold wrote a comprehensive summary of this history in his 1985 book A lot about the birth of personal computers and many of the people I just mentioned TFT went a bit quiet in the 90’s, not a lot written

Around 2013 Bret Victor starts doing talks For the newer generation his work becomes canon. The Humane Representation of Thought, and Media for Thinking the Unthinkable between 2013-2015

All of these programmers are pushing this idea that we need to be building new tools for thought. And most of the people listening to them or reading their work are computer people. And so we end up assuming the problem is a computer-shaped problem.

Another interesting thing about the way these programmers are talking about “tools for thought” is that they’re actually talking about a medium for thought, not a tool.

For philosophical clarity, these two are different

Mediums are a material we use to communicate ideas: film, painting, writing, photography

Whereas tools enable specific workflows.

You don’t communicate through a hammer or agile. They’re part of a larger context – the medium.

I think it’s important to be clear on the different between these two, but when we talk about “tools for thought” as a category of things humans have used to expand their thinking, both are part of that picture

I think its more accurate to say the thinkers I referenced so far are talking about a computational medium for thought.

Where the computer-ness and medium-ness are critical elements

It should be CMFT, rather than TFT

Also clear than CMFT is a subset of TFT If TFT is all the tools and mediums that expand the range of human thought, this community is specifically talking about the computer medium type

So what’s so special about computers?

I think this is best answered in this Alan Kay quote

Encoding cultural practices in computation reconfigures, expands, and transforms them

The meta-medium changes the practice

We have to acknowledge there’s something very different about computing - in its ability to create representations of the world that we can explore and manipulate, freed from many of the constraints of physical materials.

If we think about a Zettelkasten, which is a form of note-taking I mentioned earlier.

I think it would be slightly absurd to deny the fact a digital zettelkasten or note-taking system has very different affordance to a physical zettelkasten. Even if they’re based on the same cultural practices.

The computational version changes what we’re able to do. Clearly expands our capacity in some way

But even though computing expands some of the practices in new ways, we shouldn’t forget we’re still recreating cultural practices that first exist outside computing.

Maps and spreadsheets and zettelkastens all existed before computers.

Computing as an amplifier or enabler

At the moment there is a LOT of staring, hard at the computer. Thinking about what the computer can do. Thinking about software and interfaces and novel ways to store and retrieve data Which is necessary… to some extent. But there’s not a lot of attention put on the cultural practices we’re trying to represent

We’re currently enabling a fairly small range of activities.

Primarily:

• capturing information (and much debate over which specific format to capture it in and where that data lives)
• Organising information. This includes linking, tagging, categorising it.
• Retrieving information. This means search, and being able to find the right information at the right time. Having relevant information appear when you need it
• Remembering information. These are the spaced repetition systems.

• Searching for new information.
• Synthesising ideas.
• Generating ideas.

I’d argue we’re not very good at these last three at the moment, but that’s primarily a lot of what people are trying to get AI and language models to do. I’ll come back to AI in a bit.

These actives all serve a particular flavour of knowledge work.

Work that is based around research and written output with a textual focus. Helping people write documents and code and presentation slides.

Good for academics, social scientists, strategists, journalists, teachers, consultants, VC, Twittering thought leaders, and (unsurprisingly) programmers.

We’re mostly designing for ourselves.

At the moment, it feels like we’re not exploring computational mediums for thought, as much as we’re exploring a subset of it I’m going to call… Computational mediums for White-collar knowledge work. Of CMfWKW for convenience

Clear we need to expand our view of this field a little.

The big question is how do we help the tools for thought community diversify?

How do we help move the community beyond computers for text-based white-collar knowledge work?

I have three overarching ideas of how we can start to do this.

If we’re trying to look at what enables novel kinds of human thought, to me that’s an anthropological and cognitive science question, not a computing question.

At the moment most people in this space are trying innovate without learning much about the world outside of software, or the even outside the world of cool hyperlinks in web apps.

I think the field would benefit from a much wider lens on the problem.

Specifically, theories like embodied cognition, extended mind theory, and looking at examples of cross-cultural ways of thinking from ethnographic research could add a lot to our conception of tools for thought.

To name just a few, we could look at the work of philosophers like Andy Clark, who’s written extensively about about how human cognition interacts with the environments around us.

This book “Natural Born Cyborgs” might look a bit dated, but all the content inside it is still extremely prescient and relevant. It’s not his only book on the subject, but I think it’s one of the more accessible ones.

In it, Clark presents the extended mind hypothesis that says we don’t think inside of our brains, but we think in the relationship between our brains, our bodies and the world around us.

George Lakoff’s research on embodied metaphors as the foundation of all conceptual thought

This book is short, it’s pretty readable, and it will make you rethink the way you understand how abstract thought is constructed.

Lakoff and Johnson argue that everything is grounded in our embodied experience of the world, which certainly has implications for AI, but again I will not get into that now.

Similarly, Barbara Tversky has done an extensive amount of research on how thinking happens through the body and through action and motion in the body.

Both this book and almost any of the existing research on embodied cognition will convince you that the way humans think is deeply tied to our physical interactions with the world around us, and our understanding space, colour, light, sound, texture, and rich visual inputs.I imagine none of this is news to your research group. You know bodies are important. But not everyone does.

Some notable others include anthropologist Tim Ingold who’s done an enormous amount of work on how humans build techniques and skills in relationship with the environment they live within.

There’s also Bruno Latour, another anthropologist who worked on actor network theory. Maurice Merleau-Ponty, very famous for setting the foundations of embodied cognition. Marshall McLuhan for talking about the relationship between mediums and communication. And I’m sure you all know, Don Norman came up with the idea of cognitive artefacts.

This is not a comprehensive list, but a good starting point. I think one of the most useful things we could as the Tools for Thought community is to spread around the knowledge that these thinkers have established and use it more as the foundation of our work.

And the reason I think we’re not doing that at the moment is because I look around and have to ask, where are all the embodied visual, spatial, and aural tools for thought?

Professions that primarily work in non-textual mediums are unlikely to find the “tools for thought” we’re currently offering as profoundly helpful.

We of course of have software that serves some of these disciplines very well; spreadsheets for mathematics, 3D modelling for biologists, data science, photoshop for painters, logic pro for musicians. But strangely these don’t show up a lot in the tools for thought community.

To give you some examples of what I mean, I would consider AlphaFold from DeepMind, a good example of non-textual, spatial tool for thought.

Jacob Collier. Grammy-winning musician showing an extraordinarily complex breakdown of his song Moon River in logic pro.

Over an hour, but an exceptional example a computation medium enabling wildly different kinds of musical thought.

New image generators for visual brainstorming and exploring visual culture

These are images from Midjourney. Image generators are controversial for other reasons like copyright, but as a creative thinking tool, I think they’re worth exploring.

We still have to figure out how to get more fine grained control over the outputs. But its very promising.

VR obviously exploring embodied cognition side of things

Prisms VR for spatial reasoning with maths

Prisms VR for spatial reasoning with maths

On that note… It’s funny to be in the current moment of the AI hype bubble, where the professed goal of burning all these GPUs is to invent thinking machines.

People hope to make these machines reason and problem solve better than humans can and solve all our complex problems.

And yet we’ve gone about it by an intensely disembodied and textual technique, by literally only dumping text into these models, so that all they know is text.

Language models – the primary kind of AI most people are putting the most stock in – lack an understand of bodies, objects, and environments interacting in space and over time. Which the literature shows is core to the way humans think.

That doesn’t mean they won’t produce useful tools for us, but they will not recreate anything resembling human thinking and problem solving. At least not without giving them some way to understand embodiment and motion and interaction with the physical world.

So just consider that next time someone tells you that GPT-5 is going to solve climate change.

The second way to diversify tools for thought is to build complementary, not competitive artefacts

This theory comes from David Krakauer of the Santa Fe Institute.

Competitive artefacts do the cognitive work for you and hide their workings. So when you take the tool away, you’ve actually diminished your skills because you haven’t been practicing it. They compete with us in a way.

While complementary artefacts teach you new cognitive skills through using the tool. They complement your abilities and they make you more capable even when you don’t have the tool with you.

A good example of this is Google Maps versus a compass

Using Google Maps doesn’t make you better at navigating the world around you. It tells you exactly where to go without relating it to local landmarks or the position of the sun or helping you develop a sense of direction in the world. If you take Google Maps away, none of us are better at navigating our cities.

Whereas a compass makes you look up more, makes you pay attention to the physical world around you, and teaches you how to relate the direction you’re in to the physical world you’re seeing via landmarks like the sun. It teaches you navigation skills that would remain if we took the compass away.

One prescient example at the moment is coding assistants like Copilot.

I think used in some ways these could be complementary, but the vast majority of uses I’ve seen of them are actually competitive. They’re not helping you learn how to solve programming problems yourself, especially if you’re not reading the code that they output.

It would be interesting to try to redesign copilot and focus a making it a complementary tool that slowly teaches you programmatic thinking versus doing things like applying large code chunks to files or rewriting entire files for you.

Third thing I think is missing is the whole mission of this talk

I think we need to pay much more attention to cultural practices, not computational objects.

There are tons of pre-existing, super effective cultural practices that make us better at thinking. And we should be considering how we could enable more people to use them. Perhaps with computers, but perhaps not.

Just to give you a few examples of some under explored ones:

• Spatial memory. We have extremely effective memories when we relate facts to physical space. I haven’t seen that much exploration of designing tools to help people build memory palaces.
• Classical rhetoric. This is how the ancient Greeks and Romans developed convincing arguments. We could build tools that teach us how to use ethos, pathos, and logos. And bring in more alliteration and antithesis and analogies.
• Critical thinking skills. This is a huge category. What tools do we have that help us question and interrogate claims we come across? And get us to question our assumptions and prove ourselves wrong? And point out our unconscious biases and logical fallacies?
• Statistics and statistical manipulation. We all famously know statistics are mostly lies. But being able to understand and interpret statistics is an incredibly powerful tool to understand big data. And I think most of us, me included, are pretty illiterate in statistics. I would love tools that help me correctly interpret and put into context data that I encounter on a daily basis.
• Information literacy skills. Obviously shown to be pretty low across the population. And getting worse as AI Slop fills every single online platform. Do any of your current tools help thoroughly vet the sources and legitimacy of information you encounter throughout the day? Do they help you compare and contrast arguments you read and come to more informed and synthesised conclusions? No! Or at least mine don’t. If yours do, I want to know what apps you’re using.

There’s lots of places to look for more practices like these. Daniel Bennett has a bunch of good examples of these in his book “Intuition Pumps and Other Tools for Thinking”. He wasn’t explicitly referencing any of the pre-existing computational literature on TFT when he wrote this.

The way he thinks about TFT is similar to what I proposed at the beginning; cites words, numbers, maps, diagrams. But, being he philosopher, he is especially interested in “intuition pumps” which is his own term meaning reasoning moves that help you think critically about philosophical problems. These include following Rapport’s rules, Joosting, and Rathering. If you want to know what those are, you’ll have to read the book.

In other words, we should be building tools that give people new lenses on the world. That teach you how to see and think in useful new ways.

I know a lot of the examples I showed probably fall under the category of critical thinking lenses. But you could also teach people to see through an emotional lens, or see the world through the lens of complexity theory and emergence. To me, these are the most promising “tools for thought.”

Thanks very much for listening.