Pattern recognition for technical communicators

Editor's Notes

• #4: From the Lockport Union-Sun & Journal article:On Christmas morning, Lockport resident Paul Kulniszewski went to cut up an orange as part of his breakfast. … To Kulniszewski, the inside of the orange resembled Jesus Christ on a crucifix with his mother Mary below.Kulniszewski showed the orange to family members and gave pictures of it to a number of people. … As for the actual orange, Kulniszewski sealed it in a container with resin so to preserve the fruit. That way others can see the orange and it can be passed down to the next generation."I just feel I was meant to share it," Kulniszewski said. "I attend (church) regularly but I'm nothing special. It just means someone else is in control."
• #7: Note how we don‘t just determine the individual elements, but also the dividers, the punctuation, and assign meaning to them. Bold is for volume number. Regular number is pages. Regular in () is year.
• #8: ... Oneofthemosthelpfuland noble thingswecan do istomakepatternsmoreobvioustoourusersandcustomers, so theydon‘thaveto do thatworkas well - on top whattheyweretryingto do whentheycameuserassistance...
• #9: If we understand better what happens here and how, we can apply these processes more efficiently. And we can avoid some of the risks. We can hot-wire our understanding and hack our writing.
• #10: Pattern recognition occurs unconsciously when we encounter more or less similar items in proximity, in space or in time.Consider this example: Suppose you walk through the country and encounter cows and horses in the space of a pasture. Looking closely, you can distinguish two different kinds of animals by their horns, hooves (even- vs. odd-toed ungulates) and tails, even if the cows are different in size and color, and so are the horses.
• #11: We start off with a bundle of data and we have to quickly make sense of it. Things that aren’t close together might not be related to each other.
• #12: We could quickly say “oh, look at all this stuff together in the middle — that can all be one thing” …
• #13: … but clearly there are more complicated things going on here. We can segment according to similarity, which might mean on the basis of shape …
• #14: … or might be about shading.
• #18: … we accumulate evidence over time.
• #19: … and if we encounter that evidence more spaced out over time, it takes us longer to detect a pattern. We see this in neuroscience, at the level of individual neurons, and we also see it in human behavior. If you’re studying a second language and only look at your materials once a week, or once a month, it will take you a lot longer to internalize the rules of grammar and build a vocabulary, because the rate at which you are accumulating evidence is slower, and the incipient connections in your brain grow cold, if you like, when they’re not regularly maintained.While we’re coming at this from a tech authoring perspective, it’s also worth thinking about the end user here. If they can’t collect much evidence, or it only comes to them in dribs and drabs over time, they might not be able to discern the patterns in your documentation/user assistance/etc.
• #20: You can see this happening in the brain’s primary visual cortex as we recognize visual patterns, building up fragments to become lines, surfaces, and objects.Pattern recognition happens as the third stage of a bottom-up process of human perception, after sensation and perceptual organization (Zimbardo 262, 294). We receive a sensation from the outside world, for example, we see white, tan, brown and black masses on a floor of green. We organize the sensation perceptually: We take apart the image to distinguish standing or slowly moving animals from each other and from the pasture. (The brain can form “object representations that are invariant with respect to the dramatic fluctuations occurring on the [eye’s] retina”, suggests Olshausen, 4716.)Pattern recognition: We analyze the animals to identify similarities and differences by which to sort, order and label them. Thus we distinguish cows from horses. That is only half the story, however.
• #21: I guess most of us would recognize this as a chair. Even though most of us probably have never seen this particular chair.The reason we recognize it is the second approach in pattern recognition.
• #22: This is the shortcut approach when we can trust our knowledge of systems and context.We’ve evolved language in order to communicate the abstract rules conferred by experience. In other words, if I learn a rule, I can tell you about it and now you know the rule too.Professionally, having done lots of feature analysis is what counts as experience. In tech comm, for example, it makes the difference between an experienced tech communicator who quickly realizes how to separate information into concept and procedural topics.Unless she has similar experience from remembering hundreds of past topics… One particular way of top-down processing is called “template matching” where you store a generic idea of what a chair, a horse and a cow looks like in your memory as a template and match it to something you haven’t seen before.
• #23: Here’s a quick contrasting view of the two approaches to illustrate the differences.We apply both processes, switching back and forth really fast and without deliberate choice.But do they always work? Let’s look at another example…
• #24: It’s art! It’s in the museum of modern art in Frankfurt, where I’m from. And the title is not helping!But really: What IS it?It is also a chair. More precisely, it is…
• #25: It’s part of a chair! It’s a chair by Arne Jacobsen.
• #26: It TOTALLY emphasizes context – and if the context is modern art, all bets are off, and you don’t know WHAT you’re looking at!Now for the practical application.
• #27: There are (at least) five ways to apply pattern recognition in tech comm. And my claim is we can tweak how we work and what we produce if we know about pattern recognition and apply its two strategies of bottom-up and top-down processing consciously.None of these ideas are new or revolutionary. And they can’t be, because we use them anyway; they are part of our mental inventory. The point is to understand these strategies better and to use them deliberately, for faster and more effective writing.We’ll go through each of the stages now.
• #28: You have incomplete, outdated documentation plus a couple of SMEs who vaguely remember additional bits and pieces.Tease out all similarities until you arrive at segments.
• #29: Test top-down, check if your segments are reliable and complete: - Are there concept topics to back up procedural topics?- Is there information about exception handling?
• #30: You know you have to start writing now, or you’ll never make it by the delivery deadline. Then it’s a safe bet to start with the segments that have most elements and the most similar elements in them. Describe what hangs together first. Trust “safety in numbers” and start with topics that have the most buddies and then build outwards.
• #34: Group topics, e.g., by type: concepts vs. task vs. reference Structure each type in a recognizable, reliable way, e.g., in tasks:IntroductionPrerequisitesActual stepsExpected resultsException handlingIn headings, consider usingNoun phrases for concept topicsImperatives for task topicsGerunds for processes which are parents of several tasksThis is basically thinly veiled praise for what’s known as “parallelism”.The cool thing is that it works, even without explanation, even if most users never spell out the underlying rules.
• #35: Resisting patterns: Commas or spelling in English, articles and plurals in German - so not everything can be understood with pattern recognition. Neither in languages, nor in tech comm.Contexts, for example, national cultures: Germans like top-down; Japanese prefer bottom-up. - In art, all bets are off… Don’t forget that our way of creating documentation is very different from how customers use it. And it’s our job to make it easy to use…!