Play in Predictive Minds
Andersen, Kiverstein, Miller & Roepstorff (2022) published “Play in Predictive Minds: A cognitive theory of play” in Psychological Review. Anderson et al. begin by saying, “In this article, we argue that a predictive processing framework (PP) may provide elements for a proximate model of play in children and adults. We propose that play is a behavior in which the agent, in contexts of freedom from the demands of certain competing cognitive systems, deliberately seeks out or creates surprising situations that gravitate toward sweet-spots of relative complexity with the goal of resolving surprise. We further propose that play is experientially associated with a feel-good quality because the agent is reducing significant levels of prediction error (i.e., surprise) faster than expected.”
They review previous notions about the purpose of play, e.g., practicing needed motor and communication skills as an evolutionary strategy or exploring the world with curious minds. They note, “In most of these cognitive accounts of play, the explanation for why children play is the same: Children play because playing is fun and rewarding.” After discussing notions that play is intrinsically rewarding, they cite Chu & Schulz (2020) who observe that, in play, children often create imaginary problems for themselves they have no obvious need to solve, yet children will expend a good deal of energy with no obvious payoff for doing so. Chu and Schulz contend that “children’s propensity to adopt idiosyncratic goals may be what distinctively human play is all about” (p. 327).
Here’s more:
In its most ambitious form, the predictive processing framework (PP) attempts to explain perception, action, emotion, cognition, and their intertwined relationships via a single mechanism of prediction error minimization, whereby the brain attempts to reduce the mismatch between how it predicts the world to be and how the world actually is (Clark, 2013; Friston, 2010; Hohwy, 2013). We argue that the predictive processing framework may provide elements for a proximate cognitive theory of play that can help explain the widespread occurrence of play across form, age, and context. The universal occurrence of play across such a wide range of behavioral forms (e.g., playful object handling, playful running, playful eating, and so on) could be taken to suggest that play may be linked to domain-general cognitive processes rather than to highly specific tasks and domains. . . . Utilizing this framework, we propose that play is a behavior in which the agent, in contexts of freedom from the demands of certain competing cognitive systems, deliberately seeks out or creates surprising situations that gravitate toward sweet-spots of relative complexity with the goal of resolving surprise. We further propose that play is experientially associated with a feel-good quality because the agent is reducing significant levels of prediction error (i.e., surprise) faster than expected.
They describe numerous studies of toddlers and preschoolers, adding:
In sum, belief-violation, ambiguity and novelty provide children with learning opportunities to improve their intuitive theories of the world and its causal workings. Children often sample information selectively when they play with the aim of updating their beliefs to resolve novel or surprising observations.
In a section titled “The Right Surprise,” they describe the Goldilocks principle noting that, “Several lines of research have suggested that both children and adults in general prefer stimuli which somehow hit what might be described as a ‘sweet spot’ of surprise (e.g., Bloom, 2010, 2020; Dember & Earl, 1957; McCall & McGhee, 1977). Such sweet spots are typified by “only moderate differences between any given stimuli and the observer’s prior knowledge.”
They conclude:
The theory explains why play is fun and why play so often is characterized by just-right levels of relative complexity. Importantly, the role of valence in this account brings novel insights to the predictive processing community as well, by highlighting a new contribution of niche construction to error minimization. In play, the agent purposefully creates and resolves error, and in so doing finds their way to better than expected policies for reducing error going forward. In sum, the predictive processing account may offer a valuable proximate model of play, and help isolate key mechanisms and variables underlying one of the most universal yet open-ended behavioral categories. As a consequence, it may help to bring play research where it belongs; at the center of developmental research.
I have mentioned before the importance of play. This work is relevant to adverse effects of the pandemic on children’s ability to engage in active play. It may also be important in understanding the mixed data on video game play.