A tension map is a map of ideas. It’s a little different than, say, a diagram explaining how a sceptic system works, or a diagram saying how to model a CPU.

Rather, a tension map is for things that are way more complicated. Things that are so complicated, that you couldn’t dream of mapping them.

For example, “make a map of all the words in the dictionary.”

You’d be crushed. And you’d think that there’s no way of mapping them all out there.

Some things have obvious and clear structure: For example, you could map a table next to a chair next to a footstool next to a sofa, and so on. Some things are fairly easy.

But the next thing you know, you’re labeling the stable island that is a living room, and then you realize, “Oh wait, chairs and tables can go into the kitchen area, as well.” But you’ve already mapped the table and the chair and so on in the Living Room. Horrors!

Things get worse when you consider that offices also feature tables and chairs.

If this was the extent of things, we could get away with being clever, and describing the abstract space of human body positional needs,and shuffling our tables and chairs into that abstract space. Then connect that abstract space with the living room, and the nearby kitchen, and those two a little ways further with the office.

Things get much much more complicated than this, though.

But perhaps you can see the way this is going.

This, incidentally, is how we find ourselves in the space of the Tension Map.

The experience of a Tension is the only way that I know of, to resolve a Paradox, that you cannot, for either limits of structure or time, resolve. When we are talking about constructing maps, our dilemma is structure, rather than (we would hope) time. “Given infinite time, make a nice 2D map of this. Go.” We now have a structural problem: We cannot neatly describe the thing on a sheet of paper.

Making A Tension Map

I first described this process in my book, How to Make a Complete Map of Every Thought You Think. I’m pretty sure that’s a much more detailed explanation than what I’m going to give right here, right now. Feel free to refer to there, should you need more detail than what I say right here. I also described this process (I think, if I remember right:) in the book Mind Performance Hacks. There’s a chapter that O’Reilly wrote a nice little check out to me for writing, on mapping your thoughts, and so on. Hey, I’ve been paid to write something! How cool is that? At any rate: Feel free to ask a question or whatever. Regardless.

Tension map. Making a tension map.

You take all the ideas you want to map out, and you lay them out in a list. You then start putting the ideas out onto paper. You make them big or small, based, subjectively, on how you intuitively feel about it. “Is this a big idea?” “Is this a small idea?” Sometimes you find themes to your list of ideas, that weren’t in the list of ideas themselves. Feel free to just make a new “idea” or word on the spot, and label it: “This is the general theme that I detect here.”

It doesn’t matter if it’s “real” or not. It doesn’t matter if it’s just a construction, and could have been done another way. It doesn’t matter that it’s essentially MetaPhysics.All thoughts are like this, anyways. No thought is equivalent to the real thing. Just don’t worry about it.

Just feel out the tensions, and write out the thoughts. It’s Dumbledore’s Pensieve. You just put them down.

You put out a thought, figure roughly how big it is, and put it down.

Then you pull out the next thought. Now you have two things. You choose a relative style, a relative font, a color, whatever, and then a position. You figure out how close or how far you think that the thing should be, compared to the other thing. How far? “As far as the tensions demand.” If the thought intuitively feels like it belongs far away, then put it far away. If the thought intuitively feels like it belongs close, then put it close.

Pull out your next thought, your third thought. Where should it go? It’s the same thing. Intuit it. If we first pulled out “chair,” and then we put “table” nearby, and now we pulled out “salt & pepper shaker,” put it closer to the table, and further from the table. Put it at a right angle. If you’re doing this in 3D, hell, just put the salt & pepper shaker on top of the table, just as it is, in the material world. The trivial case, the “zero case,” of the “tension map” is zero tension, because we’re just replicating whatever we see in the material world. No need to go putting in tensions, when you don’t have to.

If we’re doing this in 2D though, you can’t put it on top of the table, because there is no “on top.” You might want to put it orthogonal to the table-chair axis, or you might want to repulse it away from the chair a little bit, but proximate the table, on the other side, and away. But try to avoid the exact opposite side, because that communicates, “This is on the exact opposite side of things, from the chair.” Unless you want to communicate that table salt is the polar opposite of a chair, stay away from that space!

Now we pull out “the family dog.” Oh, where does this go? It should perhaps triangulate with the chair, and the table, (and the proximate salt shaker.) The family dog should perhaps be further away from the table and the chair, though, because tables and chairs are bonded by the familiarity as human utensils, but the family dog is not used in the process of eating (hopefully not) or chairs. (But if you ever eat dog, on the table, by all means, put it slightly closer to the table, than the chair.)

And so on, and so forth. You just keep pulling things out of your list, and positioning them onto the Tension map. No doubt, you’ll probably need to rearrange things every now and then, and perhaps even in drastic ways. You may need to do total rearrangements. You may even want to (horror of horrors) list one thing twice, in two separate maps. This is the beginning of bifurcation, though. Whether you want to do that or not, depends on the situation, of course. Just know what you’re doing, know what you’re paying, and know what you’re getting.

This is, basically, how you make what I am calling, a “tension map.”

You can use it to map things that many people think are impossible to map.

But What Good Is It?

“What are the benefits of producing a map? Wouldn’t it be just better to go all the way, and simply resort to language? What’s the value in a visual dictionary of all words, or worse, all concepts?”

Good question. I often times doubt the utility of tension maps, myself. Some are perfectly useless- you’d be better off just making a language, and treating every individual word as if it’s just “it’s own thing,” rather than trying to draw maps and diagrams of stuff.

I suppose part of the value of the tension map is that, there are very often sensible substructures of a given tension structure. Sometimes, there are so many sensible substructures, and they have very little overlap, it’s better to just make a collection of: “Several trivial tension maps (that is, just several ordinary sensible maps without much tension,) and a tension map that connects them all together at a high level. Traditionally, this is what we call just: “a table of contents.”

The value is in being able to see the connections, and feel out the tensions.

Looking at someone else’s tension map can be perplexing: You have to ask yourself, “Gee, why did they structure it like this?” But in many ways, that’s exactly the point. Maps aren’t read left-to-right, top to bottom, like a page. Maps are read in whatever order, depending on what you need to get done. You’re trying to get from Seattle to Santa Cruz, and you engage in a conversation with the map. You go, “Okay, where’s Seattle? Where’s Santa Cruz?” Once you have figured out where those are, you have a number of other conversations with the map: “Map, what ways can get me from Seattle to Santa Cruz? What different roads can I take? Why ways should I not think about? If there’s going to be snow, what do I want to avoid?” And so on, and so forth: The map has a power to answer those questions, that no textual spelling out could possibly answer.

(Chalk another one up for throwing out the bogus notion of “visual learners” vs. “aural learners.” The simple fact is, visual descriptions and aural descriptions are totally different things, and what we’re really talking about is whether someone trusts the speaker to tell one what they need to know, or whether they’d rather hold a map that they can ask questions of, and get answers from, but at the cost of actually having to ask those questions. The proposed “visual / aural” distinction is otherwise bogus, as far as I can tell.)

Now, does this explanation about the utility of maps of physical space really carry over to our description of tension maps that are about ideas?

Yes it does, I strongly believe. It won’t work for all tension maps, but it’ll work for many of them.

The thing is, you look at the tension map of ideas, and you see many levels of structure. Granted, it’s not strong, (like a material map,) but it does communicate some intuitive realities.

You can look at the map, and say, “What ideas are roughly in the proximity of this idea, and what ideas are roughly in the proximity of those ideas?” There will likely be some ideas that should be close, but that, for pragmatic reasons, are actually on the other side of the map. (Stronger tension pull to those other ideas, than this particular idea, given the present superstructure of the page.) But you’ll get at least some idea of how this idea connects with others, in a way that the author considered important.

Just don’t take it too literally, unless the author told you, “Take this literally.” To read a tension map, to use a tension map, you have to accept ambiguity, from the get go, and be able to operate in the ambiguous situation.

Fortunately, we’ve all been doing that since the time we’ve been born, so it shouldn’t be too hard to get the hang of.


It occurs to me in writing this: “I wonder if this is connected with what HelmutLeitner is always trying to talk about, with ChristopherAlexander and his living patterns of building architecture, ..?”

It’s always made more sense to me to talk about ideas and ways of living, rather than building architectures, because I hardly ever think about building architectures: I practically live online, in the NooSphere. All the buildings we make are made out of ideas, not concrete and mortar, for bodies.

But I suspect that this is connected. (Helmut? Did you make it through the above?)

My apologies for not being a skilled writer. I do like the manga artists, though: Drawing crudely, in black and white, to get greater coverage and bigger ideas, rather than going for a mere 24 pages of glossy color panels, beautifully inked. But it’s worse than that; I just sketch, and I don’t even know how to neatly ink, much less begin to make characters proportions right, and get the ink to go just so, and to fill it in with color. I really admire SunirShah.

Tell me if a part needs more attention, and I believe I can fill it in better.

I should probably upload some pictures for this article…

It should also be noted that if it’s easier for you to specify closeness numerically in a table than visually, there are computer algorithms (and, I believe, programs) which take as input a list of nodes and a list of distances between nodes, and which yield as output a 2D layout of the nodes such that the geometric distances between the nodes in the 2D layout is as close as possible to the distances you gave in your table. I believe there are also algorithms that could produce a 3D layout in a similar manner, although for obvious reasons those aren’t as popular yet.

These algorithms are generally referred to as “dimensionality reduction”, more generally “data visualization”, and one of the simpler ones (which, I think, produces a “best 2D layout” in the least-squares sense) is multidimensional scaling (MDS). I believe you can get GraphViz to do MDS layout by asking it to layout an undirected graph whose edges have edge weights.

I have high hopes for the technology, one day.

But, there are weaknesses, that cause me to prefer the incredible pain of doing it by hand.

For example:

  • If a node X is diametrically opposite a node Z on the other side of node Y, it implies (communicates to a human) that there is an opposites relationship between the two things. If this is not what is meant to be communicated, then it’s best to put the second thing slightly askew from the “direct opposite” position.
  • Anything that is written above and to the right of another thing, implies, (in left-to-right reading cultures,) “In the bright future.” Anything that is written below and to the right of an other thing, implies, “In the bad future, that we don’t want.” Anything that is written to the left and above, implies: “In the beautiful past,” and anything written to the left and below, implies, “In the bad old days.”
  • Hodge-podge jumbled collections written in a poor font can be used to communicate, “these are a bunch of ideas we had.” Write them in a nice font, and it implies, “these are intuited pseudo-spatial relationships between ideas.” Draw lines between things, and not between other thing, and you say, “this is structurally connected, but this is merely associated.” You can draw lines thick, wavering, with solidity or little edge points missing, … (and so on, and so forth.)

This is just a small handful of meaningful variations that come into consideration upon drawing a TensionMap.

Deducing all these things algorithmically is insane, by present computing standards. We’ll have to have Jeff Hawkin’s electronic brains working, before we can even imagine software that can present at the level of communicative subtlety (the level I care about here) that a human brain can do casually (if laboriously.)

Hm; I hope this makes my position clearer– I’ve been emailed more than a few times by people informing me that I’m a Luddite (!!) for discounting computers. ;)

One of the major challenges of our day is to collaboratively create an enormous Tension Map of all public groups., and perhaps even some “blank spots” for secret groups..!

This is, of course, a ridiculous task, because this all relies on “intuition” and subjective feeling, of course. None the less, I believe it can be made to work, by proper distributions of delegation, and a proper rules framework system. It may be possible to use CollectiveIntelligence, and release “the map,” or more likely, several maps, on a yearly or semi-yearly basis; Something like that.

You could even remix existing maps of groups and people.

SecondLife has a map, and the map works, even though it’s arbitrary. “So there.” We would want something, I think, that was based in more than just “when you bought the property,” because it has to do with ideas and so on.

Deducing all these things algorithmically is insane, by present computing standards.

If there is someone like you around to catalog the various stylistic mechanisms, like you have started to do in that comment, then computers could be programmed to make use of them (with a “markup source” that is fed into a “diagram compiler”, just like wiki source). Most people who would do it by hand wouldn’t be creative enough to invent new stylistic mechanisms anyhow, and would even only learn to do the “old” ones by reading lists like the one you had in that comment. People like you who are into diagramming enough to actually invent new stylistic conventions could still do so ()by hand), but i think that after the first 20 most common conventions or so are cataloged, the vast majority of people who will be making diagrams won’t lose much by using automated software which accepts “diagram source” and applies the known conventions to generate the diagram.

This is assuming that diagrams become more common than they are now, because right now a lot of the people who make diagrams online are probably “visual language enthusiasts” like yourself. But if EVERYONE makes diagrams, that would no longer be the case.

Well, you still have to measure the tensions. But I think there’s a niche for the kind of automated thing that your talking about, and that that niche would widen out as we taught the computers more ideas.

Lion, I have that Mind Hacks book you talk about above. Which part did you write in it?

Bayle took the words right out of my mouth, and you guys already covered my main thought about this. I definitely agree that drawing lines between objects on the map would not help, because, and this is articulated in SocialNetworkAnalysis schools of thought, it will tend to implant incorrect interpretations in the minds of people who view them. SocialNetworkAnalysis practicioners have done experiments where they showed incorrect maps to executives, and they get a significant portion of people responding saying something like “yes, I suspected that this is how people are connected”. So, drawing lines between things potentially puts very powerful suggestions in people’s minds.

When Bayle was talking about algorithms and computers, it made me think about Cross Impact Analysis, which is a futures studies technique that would put your root TensionMap list in a matrix, and place some kind of value for every variable in how each item relates to each other. This might be a way of creating a system that a machine can map out, eventually. But of course, it depends on people to create the variables, and a way to understand how the values relate to connectedness/closeness between objects.

Conversely, it seems like I’ve come across co-editable whiteboards that would let you place objects on them and move them around (also, SecondLife of course lets you create 3D objects and move them around :))

Oh, no; I’m talking about Mind Performance Hacks [1]. (Review on the Mind (blank) Hacks website. [2]) The author, Ron Hale-Evans, is a friend of mine, (pic of me with my Cosmic Encounters prize [3],) and he asked me to write a few chapters. I wrote a chapter on the Major system (changing numbers into phonetics, and vice versa, as a mnemonic system; ex: Chromium → “Lit up bush” → 51.996 AMU,) a chapter on pre-deleting cruft, and a chapter on mapping thoughts, such as described on this page. Or did someone else write the Major system chapter? I don’t quite remember. Regardless, that’s the one.

Eh… As for drawing lines; I find way too much value in drawing lines at some point, to just dismiss them altogether. If there was a study that lines were a problem in some sort of SNA practice, I would have to ask questions like: “Okay, what exactly were these SNA maps like?” There’s got to be 2,000,000,000 ways to map a SocialNetwork. If you show bad maps to people and they get bad conclusions, I wouldn’t be surprised. I do agree that lines draw powerful connections. (That, in my mind, is what they are for.) Though, not always– it’s easy to make a hierarchy of lines, and you can even say, “these lines aren’t so important” by putting a bunch in random mish mesh, you can make them curved, squiggly, dotted, colored, you can make them loop entities or wrap around a… …I mean, there’s just so much variety.

To properly evaluate a use of line, I think you have to really look at the whole systematic language of use around that line. You have to look at how the letters are being used, the fonts, the lines, the shape, what kind of variety there are, and so on. It’s like letters: It could be Spanish, English, romanized Japanese, …

The SocialMapApril?2006Image map uses lines, I think to powerful effect. People will likely get wrong ideas from looking at the map, if they don’t know who we are. That said, even if we spell out the whole map in paragraphs of text, people will likely get wrong ideas (still.) So, I’m not sure that it’s that big of a problem.

Even conventional SNA lines, I think, can work. If you say, for example, that lines mean “personal friends, for more than 5 years,” you see something. If you say, for example, that “lines have a dotted pattern, that explain when they were in close proximity and communicating,” that means another thing. (and so on.)

I would ask to see those studies, those papers, and check their methods, and I would probably end up asking them to redo their studies, this time taking X Y and Z into account. This expression of mine reminds me of RobertHorn? near perpetual battle against straw man studies, purporting to show that “diagrams don’t mean a damn thing.” Ah, right, here it is: How to Get Little or No Effect and Make No Significant Difference.

Well, what I was talking about with the lines was kind of a side note that Robert Cross found that when people saw innaccurate lines, they would still tend to believe them. So, this means that lines are awesome, and powerfully descriptive and persuasive, even when they not accurate. SNA relies heavily on lines between nodes on diagrams. So, they were definitely saying that lines are good, but they were saying be careful about showing people diagrams that are still works in progress, because those if those lines turn out to be inaccurate, they’ll imprint on people’s minds. I also agree that different types of lines can help differentiate, too. It may seem like a small thing, but I think that as more people put collaborative visualizations, or even visualizations of data aggregations, it’s worth thinking about how they are interpreted.

Also, I thought that lines on a TensionMap in particular might not work well because of all of the variables of relationship that are being represented. But then again, I can be totally wrong about that. I thought there might be too many variables to accurately represent with lines in a TensionMap diagram. I guess when I was reading your description of how to create a TensionMap, I was imagining something more like a scatter-plot of images or objects that might end up in clusters depending upon how the are connected in different ways.

Of course, lines could totally work. But, I wonder if the map is considering very many variables, if lines would end up creating huge mess, as the sometimes do when you start combining lots of different data in SocialNetworkAnalysis. You still learn from the map because of the way things are grouped, but it becomes tough to see the connections with lines if you combine many different “views”, like who talks to who, who do you get information from most frequently, who knows what, etc.

Oh! Okay. I see it. :)

Yes, yes, we can lie and misrepresent in diagrams, too. ;)

Quite. :)

(I don’t mean to accuse anybody of lying; It’s just the phrase to explicate: “It can be different than what’s shown.”)

It may seem like a small thing, but I think that as more people put collaborative visualizations, or even visualizations of data aggregations, it’s worth thinking about how they are interpreted.

I agree strongly, (and preach) what RobertHorn? says: That we are, as humanity, right now, literally building language. And not just one, but lots of them..! With dialects and creole’s and everything..! And I think it’s this very evolutionary thing. “How does this word or image stimulate you?” “Well, it makes me think X.” “Hm, that didn’t seem to work; How about this?” Meanwhile, we unconsciously repeat the patterns that we don’t even think about them, because we’re so immersed in them.

Yes. A lot of what you’ve been working on in the recent past here in CW could be an example of this emerging new language. You really don’t see people talking about VisualLanguage too much. For instance, look at the impressive tools, like InkscapeToOddmuse, and the knowledge that people have here as a group. Someone could easily say “Hey Lion, let’s try and make a TensionMap here.” And we could use the InkscapeToOddmuse setup to work on it asycnhronously over time. I, for one, am very interested in trying that out, even as an exercise. But, it could also be an application, such as mapping out CommunityWiki, at the same time.

BTW, ran across an example of your tension map for CIC (http://cic.evolutionarynexus.org/files/20061226_lion_mapofci.png), which gives me a good idea of what you’re talking about here. It is kind of like a scatter-plot. And, I really wonder if cross impact analysis might be a way of collectively generating one of these. You’d still create the map by hand, but you could do a workshop before hand where an interested group uses a matrix to create values for relationships, that could give the map-maker a way to aggregate group intelligence. Or, instead of a workshop, you could use a Delphi method where each person fills in a matrix independently, and then the map maker creates the map with a guide of the aggregated values of different people working independently.


Define external redirect: RobertHorn SocialMapApril

EditNearLinks: SunirShah ChristopherAlexander GraphViz