Learning As Pattern Recognition

Share Button

Brett Steenbarger, a professor of psychiatry at SUNY Upstate Medical University, has some learning advice for financial traders that is useful for the rest of us:

The core skill of trading is pattern recognition.
Whether the trader is visually inspecting charts or analyzing signals statistically, pattern recognition lies at the heart of trading. The trader is trying to identify shifts in demand and supply in real time and is responding to patterns that are indicative of such shifts.

Much pattern recognition is based on implicit learning.
Implicit learning occurs when people are repeatedly exposed to complex patterns and eventually internalize those, even though they cannot verbalize the rules underlying those patterns. This is how children learn language and grammar, and it is how we learn to navigate our way through complex social interactions.
Implicit learning manifests itself as a ‘feel’ for a performance activity and facilitates a rapidity of pattern recognition that would not be possible through ordinary analysis. Even system developers, who rely upon explicit signals for trading, report that their frequent exposure to data gives them a feel for which variables will be promising and which will not during their testing.
Research tells us that implicit learning only occurs after we have undergone thousands of learning trials. This is why trading competence—like competence at other performance activities such as piloting a fighter jet and chess—requires considerable practice and exposure to realistic scenarios.
Without such immersive exposure, traders never truly internalize the patterns in their markets and time frames.

Emotional, cognitive, and physical factors disrupt access to patterns we have acquired implicitly.
Once a performer has developed skills and moved along the path toward competence and expertise, psychology becomes important in sustaining consistency of performance. Many performance disruptions are caused when shifts in our cognitive, emotional, and/or physical states obscure the felt tendencies and intuitions that lie at the heart of implicit learning.
This most commonly occurs as a result of performance anxiety—our fears about the outcome of our performance interfere with the access to the knowledge and skills needed to facilitate that performance.

Engage in a structured training process.
Education—simply reading articles in magazines, websites, blogs, and books—is important, but it is not training. Training is the systematic work on oneself to build skills and hone performance. It requires constant feedback about your performance—what is working and what isn’t—and it requires a steady process of drilling skills until they become automatic. No amount of talking with a coach or counselor will substitute for the training process: not in trading, not in athletics, and not in the dramatic arts. (Read more here.)

Think for a moment about what “pattern recognition” means in your own job. For me, it means being able to recognize an interesting insight hidden within often-dense academic papers, and being able to fit that insight into the structure of things I know about learning, as well as into the structure of a blog post or a magazine article.

When I worked at Psychology Today magazine many years ago, and had to churn out dozens of news items for every issue, I actually made a list of common story-telling structures: “So you thought X was the case? Well, new research shows that actually Y is true.” “New research shows that that old bit of wisdom, X, is more true than we realized.” “Here’s something you’ve never considered: X.”

Rather than leading to formulaic, same-sounding stories, these templates gave me opportunities to showcase what was unique and different about each new study I wrote about. I was reminded of my templates when I read Made To Stick, the wonderful book by Chip and Dan Heath, which pointed out that the most effective—and creative—advertisements are those which actually work within established storytelling conventions.

This long digression brings me back to the question I posed above: What are some “templates” or commonly-occurring patterns in your own work, and how can you more precisely define and refine them? I’d love to hear your thoughts.

Share Button

7 Responses to “Learning As Pattern Recognition”

  1. One of the skills which babies develop very early and which is at the heart of the language learning they do is pattern recognition. Unfortunately this very skill is ignored by most language teaching approaches. Instead the patterns are presented to the learners as fait a compli, so the pattern recognition skills we need at all levels and areas of language acquisition are not not really called up at all. It is no wonder then that the success rates in second language learning are so poor.

  2. Marshall Kirkpatrick says:

    Fascinating and inspiring write-up!

    To answer your question, when I was a tech news journalist my most common pattern recognition was to say “this item in a stream of updates appears to reference an otherwise unknown development impacting a consumer market technology platform’s consumers or developers in a way that will cause them to have more or less freedom to innovate or consume information through innovative channels.”

    That was the day to day but more fun was the pattern recognition that pointed to new streams of data that could surface the kinds of patterns above more effectively.
    For example, we once used a machine-harvested set of data concerning the number of outbound links in the history of our website to each domain that we’d ever linked to in order to determine which domains we had linked to more than once. We eyeballed and removed any domains belonging to competing news sites, leaving a list of 1k+ vendors we had written about more than once. We then put those domains into Mechanical Turk and asked the workers there to get the RSS feeds associated with the companies behind the domains (they weren’t standardized.). Then we took the resulting RSS feeds and wrote a bot to check them every 5 mins and send any new updates from the blogs of companies we’d written about more than once into a Skype chat room all our writers watches. Then we could do the kind of pattern recognition we did on all our streams of info.

    In this case it was the recognition of the opportunity inside the original body of machine harvested outbound link counting, relative to other bodies of data we came across, that represented some pattern recognition. Doing that kind of work system building was the funnest part of the job. Of course all of that was turned off when tr company was acquired.

    Now as a startup guy looking for product-market fit, perhaps it is my job to find patterns that point towards user feedback that references an opportunity to bring together the needs of the customer as they understand them and the opportunity presented by the technology as I understand it. I’m going to have to think about this more though!

  3. EssaySnark says:

    We help applicants develop their pitch for MBA programs like Harvard and Stanford. The core of the work is assessing the common elements of a profile – basic ‘stats’ like GMAT score and undergrad GPA, plus work experience (nature and quality of work, career progression and advancement), extracurriculars like volunteering and personal interests, and demographics.

    The heuristic we apply in evaluating a profile is one we heard from a screenwriting teacher, Blake Snyder: Bschool candidates (or screenplays/movies) should be “similar, but different.” We look for qualities that show that you are similar enough to the pool of accepted students at your target school, yet different enough to stand out from the crowd. Just like with movies, if you’re too avant garde, then you’re not likely to be a mainstream success, but if you’re too same-same then you’ll probably get passed over.

    This pattern recognition thing is really interesting, thanks for this – will be thinking about it more for sure.

  4. Kathy Sierra says:

    I co-wrote a book on software development Design Patterns, but aside from the “formal” patterns, expert programmers and engineers have the notion of “bad code smell”, for the patterns that resist introspection because they were grasped at a subconscious level. For a few years I was tasked with training others by “extracting the patterns of experts and then representing them in patterns and rules for non-experts”. We actually thought that was a good idea: hey, experts have more and higher-resolution patterns, so let’s just transfer them into the heads of others. Somehow we thought patterns would work on this transfer approach even while knowing that other forms of knowledge do not.

    And of course we were so wrong. Now I use the near-opposite approach, as it seems the science suggests: to be truly useful as expert patterns, they MUST be “apprehended” by the learner himself. However, there are some patterns for patterns :) we use to help accelerate the process for novice and advanced learners to “discover” patterns:

    * high quantities of examples with very low-latency, unambiguous feedback (chicken sexing being the extreme example of this approach)

    * increasingly higher-resolution, more subtle examples, at a level of progression the learner is ready for.

    * showing/experiencing a lot of examples that LOOK different, but are actually the same at some fundamental, underlying structural level (it is this underlying structurual pattern we want them to “discover”), but at the level of resolution and subtlety they are ready for

    * showing/experiencing a lot of examples that LOOK the same, but are actually different in crucial ways that we want them to begin to distinguish.

    Dan Meyers has pointed to research that supports the idea of pattern discovery first, THEN explanations for what they have “seen” and “uncovered” second. That order matters quite a lot, though most typical edu does it exactly backwards: here is a concept, and then we’ll see examples. True, deep, patterns — the kind that become expert intuition — are not taught through direct fact and procedure instruction, but through perceptual exercises. I hope to see a lot more learning pros begin finding more patterns for helping others grasp patterns :)

    • Yes Kathy! This is great.

      If you get a chance, check out The Puzzle School (http://puzzleschool.com) – This is exactly what I’m trying to accomplish. Learning through frequent challenges, subtly increasing in complexity, with tight feedback loops, with the goal of building pattern recognition.

      So happy to hear someone with your experience trying to teach others taking this approach as well.

  5. […] had the pleasure of reading Annie Murphy Paul’s blog last week.  This particular entry was essentially about knowing your purpose.  AMP  uses the […]

  6. Very interesting – even the title of a blog post is chosen to trigger likely learned patterns (whether these represent true of false positives).

    So in an environment where pattern compliance is deliberately sought, skills evolves towards masquerade detection.

Leave a Reply