Minimal nativism

(After blogging a little less in the last few months, I'm trying out a new idea: I'm going to write a series of short posts about theoretical ideas I've been thinking about.)

Is human knowledge built using a set of of perceptual primitives combined by the statistical structure of the environment, or does it instead rest on a foundation of pre-existing, universal concepts? The question of innateness is likely the oldest and most controversial in developmental psychology (think Plato vs. Aristotle, Locke vs. Descartes). In modern developmental work, this question so bifurcates the research literature that it can often feel like scientists are playing for different "teams," with incommensurable assumptions, goals, and even methods. But these divisions have a profoundly negative effect on our science. Throughout my research career, I've bounced back and forth between research groups and even institutions that are often seen as playing on different teams from one another (even if the principals involved personally hold much more nuanced positions). Yet it seems obvious that neither has sole claim to the truth. What does a middle position look like?

One possibility is a minimal nativist position. This term is developed in Noah Goodman and Tomer Ullman's work, showing up first in a very nice paper called Learning a Theory of Causality.* In that paper, they write:

... this [work] suggests a novel take on nativism—a minimal nativism—in which strong but domain-general inference and representational resources are aided by weaker, domain-specific perceptual input analyzers.

This statement comes in the context of the authors proposal that infants' theory of causal reasoning – often considered a primary innate building block of cognition – could in principle be constructed by a probabilistic learner. But that learner would still need some starting point; in particular, here the authors' learner had access to 1) a logical language of thought and 2) some basic information about causal interventions, perhaps from the infant's innate knowledge about contact causality or the actions of social agents (these are the "input analyzers" in the quote above).


How do we generalize this idea to other parts of development? As I see it, the fundamental theses of minimal nativism are 1) that human cognition develops out of a set of evolved architectural processing streams for particular kinds of stimuli and 2) that sensory data, filtered through these processes, are combined via learning mechanisms that can create new abstractions. The hypothesis is then that these mechanisms will lead to the observed patterns in the growth of knowledge across development, which include a core of seemingly universal and evolutionarily-conserved patterns as well as some others that are more culturally variable. Let's examine these two theses.

Regarding the first thesis: In current nativist theory, infants' initial state is often referred to as their "core knowledge," and this position is taken to support a much stronger brand of content nativism. For example, even young infants are apparently able to reason about the physical world and the interactions between objects in the environment.** A minimal nativist accepts this large and compelling body of evidence but with two interpretive caveats. First, since there is every reason to suppose that these physical reasoning abilities are conserved in other primates, physical reasoning abilities could be classified as a product of the evolved primate visual system and its hierarchical organization into discrete processing streams – that is, as a property of an architecture, rather than as propositional knowledge. Second, the evidence on newborns is complex and often difficult to interpret, so we don't know how much the behaviors we observe in 3 – 6 month-olds are enabled by basic properties of the visual architecture versus by several months of visual experience.***

And regarding the second thesis: The minimal nativist position accepts the Fodorean critique of simple associationism that abstractions cannot arise from a system that doesn't have the capacity to represent those abstractions to begin with. But such a system does not need to have the abstractions themselves built in – it only needs to have A) the basic expressive or logical capacities necessary to form new abstractions (with the lambda calculus being an example of such a system) and B) a principle for deciding between these abstractions on the basis of data (Bayesian inference is one proposal for such a principle). In other words, if you assume only simple architectural biases on perceptual input, you need powerful learning systems that can create new abstract primitives – which in turn can be inputs to higher levels of learning.

What I find appealing about minimal nativism is that it allows the developmental theorist to acknowledge demonstrations of competence in early infancy and consider possible biological origins for these behaviors. Yet it also acknowledges the breadth of the gap between these early signals and the richness and generality of adults' cognition. From this perspective, perhaps the most important research challenge is figuring out how to bridge the gap.

---
(minor edits 7/12/16 evening)
* Actually, the term also shows up in Ramsey & Stich (1990), but they use it to mean "a commitment to some sort of nativism, without a clear sense of what the nature of the innate knowledge is." That seems different.
** One classic paper by Spelke here.
***  A tremendous amount about object perception is changing even in the first months of life.