Learning by noticing

From Discussion of Lucas, “Human Capital and Growth” by Sendhil Mullainathan and Learning by Noticing: Theory and Experimental Evidence in Farming by Sendhil Mullainathan et al.

Survival, prosperity, and reproduction are dependent upon productivity achieved in an environment characterized by heterogeneous rates of change, fluctuating variation and random exogenous shocks compounded by internally generated shocks (unintended consequences). There is no one-size-fits-all in terms of knowledge, behaviors and values that provides an optimal answer to all problems at all times. The habits of decision-making is the key variable for recombining knowledge, behaviors and values in different recipes to meet unexpected circumstances. A key attribute in decision-making is the need to recognize new or changed patterns of information or data.

Imagine going to a friend’s house for lunch. You watch him prepare crepes, which turn out to be delicious. The ingredients are easy to get. The steps make sense. It doesn’t take a long time. You decide to go home and try it. Our basic models of technology usage and adoption can be captured in the context of this simple example. One view of technology adoption emphasizes uncertainty over the benefits of a technology (Besley and Case 1993, 1994). If you’ve never had a crepe, you don’t know whether it is worth the trouble to try or make. Another emphasizes learning by doing (Jovanovic and Nyarko 1996; Foster and Rosenzweig 1995; Conley and Udry 2010). If you don’t know the steps of making a crepe, there’s a lot of trial, error and practice in making one. Under both of these views, this missing knowledge can be filled by watching others (as in watching your friend cook the crepe and tasting it) or by experimentation on your own.

These models miss a potentially important feature of knowledge. After seeing your friend make the crepe, you go home and repeat the steps exactly as you saw them. The result is a gooey mess. No matter how many times you follow the steps (you even wrote them down), it is far from tasty and usually inedible. The problem? What you “saw” your friend do was limited by your knowledge of cooking. You only noticed the features that you thought could matter for the end result. Yet if you know little about cooking (or about cooking crepes), you may miss essential steps. In fact, if your friend watched you attempt the recipe, he would point out errors you are making, even when you thought you were doing exactly what he did.1 What you learn about a technology depends on what you notice; we call this learning by noticing.

The implications are teased out in a slide presentation, the first document linked above. Key slides:

Learning by noticing

• Simple theory:

– Many (many) pieces of data to attend to
– Selective attention
– Beliefs drive what is attended to

• Two forms of Learning

– Learning within a mental model
– Changing the mental model

and

Rethinking Human Capital

• What is human capital?

– What you know
– What model you believe in

• Better (not more) human capital speeds up learning even on one’s own.

– It allow for better “conversations” with nature itself

• Human capital, like some physical capital, has lock in effects. Can be a strength and weakness

– ideas legislate their own borders

This line of thinking seems consistent with the observation that different value systems, different cultures (behaviors) and different portfolios of knowledge will likely have a material affect on the capacity to recognize new patterns and to learn by noticing.

I suspect that the habit of enthusiastic voluntary reading ties in here as well. That the values and behaviors which foster enthusiastic reading (curiosity, empathy, etc.) leads one to voluminous knowledge acquisition (direct and indirect) as well as encourages certain useful patterns of searching for and being attuned to nuanced patterns - hence to increased productivity. It is placing a heavy tank on a weak bridge but I see elements of support for the conclusion that enthusiastic reading improves one's capacity to both improve learning within a mental model as well as improving the capacity to step outside the model - enhancing productivity both incrementally and in a punctuated fashion.

In fact, I would go further and postulate that language, culture, and portfolio of stories would have a substantial impact on the capacity of individuals to notice, learn, and improve. This ties to Stuart Kauffman's Adjacent Possible idea - incremental advances in productivity or knowledge are contingent on the existing foundation of knowledge. You can't go straight from the concept of a wheel to a Ferrari; you necessarily have to go through several intermediate stages first.

This would explain why there is such a huge variance in child preparedness when they first show up to school at age five (a variance of some three years). Children from privileged backgrounds with significant interaction with both parents and a rich environment in both speech and reading have already acquired a much larger vocabulary (adjacent possible), priming their pump for much greater knowledge acquisition. They have likely also acquired habits of pattern recognition and noticing which serve as force multipliers for both knowledge acquisition as well as knowledge creation. Extensive reading has also likely enfused them with habits of imagination which yet further enhance their capacity for productivity (via stepping outside the model).

That's a huge layer cake of speculation but it provides a framework for finding evidence.