Tuesday 26 January 2016

Information is typically dense and continuous (A Purple Peril)

Optic flow is everywhere, all the time (same with other energy arrays, like the acoustic array). We depend on this fact deeply. When we are cast adrift from information, our behaviour quickly accumulates errors and strays in often disastrous ways. One example is the case of friction, which doesn't exist until two surfaces are in contact and therefore does not create information about itself that is available ahead of that contact. In another example, when cut off from landmarks, people walk in huge circles, getting seriously lost and confused; Souman, Frissen, Sreenivasa & Ernst, 2009). A simple version of this is the game of walking with your eyes closed; you quickly lose all confidence about where you are and what's happening and it's actually very difficult to make yourself walk at normal speed. 

The Perilous proposal is that behaviour emerges in real time, as a function of the current flow of information, and that this flow is typically dense and continuous, not intermittent. I will illustrate this with an example of two designed sets of instructions for navigating though a building, where the dense information set leads to better, more stable behaviour.

Directing people through environments is a common design problem. Designed solutions work best when they work with perception, rather than in competition (which of course implies designers need to know about perception; see Andrew Hinton's excellent book Understanding Context for much more on this). 

Typical solutions involve providing intermittent signage; the assumption is that you only need information at decision points such as intersections and will be able to happily move from one such point to the next with no additional help. My personal experience is that as soon as I'm out of view of the sign I can feel the panic rising, as if I had just closed my eyes (similar to the problem I have with staggered and especially puffin pedestrian crossings). 

Leeds Beckett University has a bunch of construction on the go right now, and the simplest route from the Portland building to the Rose Bowl is currently blocked. The alternative route winds through multiple corridors and down two flights of stairs. There are two sets of instructions: intermittent and dense.

The intermittent signage is fairly standard: signs with arrows at various key points (see Figure 1)
Figure 1. Intermittent signage
This signage is fairly useless. It contains little information, and I lose access to that information as soon as I move away.

Luckily, they also now have this version: (Figure 2)
Figure 2. Dense signage
All along the path, there are footprints. The ones heading towards Portland are labelled 'University Buildings' and the ones leading out are labelled 'Exit'. But more usefully than words, the footprints are oriented in a particular way, they follow lanes that stick to the left (this is the UK, after all) and they are densely packed for most of the way. I am in perceptual contact with these at all times through the building. They do get less dense as the trail nears the end, which signals that the trail is coming to an end and you will shortly no longer need the help. See the embedded video for a walk through:

(About 10s in, a guy walks by in the 'wrong lane'. This is a nice example of a desire line overcoming a designed solution - he's about to take a 180° turn and the wide approach provides the smoothest (least change in acceleration) path. I love desire lines :)

Another example of this style of design is the new terminal at Atlanta Airport, not usually known as a building you want to spend any time in. Listen to this 99% Invisible podcast to hear all about how the design simply and effectively continuously guides people through to where they have to go without any signs at all. 

This kind of designed, dense information flow works well because it works with perception. The dense signage is much more like the continuous information flow vision typically works with than the intermittent signage. We control our movements in real time, in the moment, as a function of the current state of the information we are embedded in. Anything less and the system becomes less stable, which we experience as frustration, hesitation and error. 


  1. I agree, with this fully 'behaviour emerges in real time, as a function of the current flow of information, and that this flow is typically dense and continuous, not intermittent' and this is beautifully consistent with what Powers has be proposing since the 1950s. However, unlike Gibson, Powers (a) produced a functional architecture that can implement this (b) realised that the behaviour is emergent because it controls perception ('the current flow of information) via feedback functions in the environment, to counteract disturbances that match the information flow with reference values for that flow. Only by including reference values (in a hierarchical system), feedback functions, and disturbances can the emergence be accounted for, and replicated reliably. Maybe you have the equivalent of these components in your apparently 'closed loop' model, or maybe not? Please explain how to build a functional system that engages in the behaviours you describe. Again - we are on the same page fighting the behavioural and cognitive assumptions on some of this - but you need to go further. And it probably won't involve jettisoning your existing models, data, etc - just seeing them from a wider, working, perspective....

  2. Sorry that should say 'counteract disturbances to match...'

  3. Your “perilous proposal” (which is not really that perilous since it is the basic assumption of S-R and cognitive psychology) goes over the edge because it looks at only one side of the closed loop relationship between behavior and the flow of information. One side is what you describe; the input-output relationship where output (behavior) is a function of input (the flow of information). The other side, which you ignore, is the output-input relationship where input (the flow of information) is a function of output (behavior). These two relationships are occurring simultaneously: what you are doing (output) depends on what is happening to you (input) while, at the same time, what is happening to you depends on what you are doing. So as an attractive person walks by the flow of information (movement of the image of the person) affects your head movement while, at the same time, your head movement affects how much the information (the image) flows.

    So while it is true that behavior is a function of the current flow of information, it is simultaneously true that the current flow of information is a function of behavior. Indeed, it is often true that information, such as the image of a stationary object, only “flows” when we are moving relative to it; and it is also often true that information, such as the image of a moving object (such as the attractive passerby), remains stationary when we move relative to it.

    When the simultaneous input-output and output-input relationships that make up the closed-loop in which behavior occurs are properly taken into account we see that behavior is a process of control -- and that what is controlled is perceptual input – the flow of information in your example. The fact that input information is controlled by (and not in control of) behavior is demonstrated by what is actually the basic PCT demonstration of principle: the compensatory tracking task. The first two demos at my website, http://www.mindreadings.com/demos.htm, the ones called “Nature of Control” and “S-R versus Control”, demonstrate this fact in a rather dramatic way. They show that the only information that could guide the behavior that is keeping the cursor on the target has no relationship at all to that behavior.

    The fact that behavior occurs in a closed negative feedback loop – one where input causes output at the same time as output causes input – means that behavior is a process of control and that what is controlled is perception (informational input). This was the observation that led to the development of PCT and it means that understanding behavior is a matter of understanding what perceptual aspects of the environment organisms control and how they control it. Research methods aimed at discovering controlled perceptual (informational) variables are described in my collection of papers entitled “Doing Research on Purpose”.

    Best regards


  4. Hi Shiya

    As Robert DiNiro once famously said to himself in the movie "Taxi Driver", "Are you talking' to me"? About the website, that is? If so, feel free to contact me at the email address at my website: www.mindreadings.com