Ecological Dynamics Approach Informs the Use of Artificial Intelligence in Sport
Ecological dynamics construes that goal-directed behaviours in sports emanate from the hard-assembled (physical) and soft-assembled (informational) links between performers and their performance environments (Kugler&Turvey, 1987). This idea implies that ecophysical variables (Araujo, Davids, & Renshaw, 2020) provide the most appropriate starting point to capture intelligent behaviour (see Chapter 3). These are variables that express the fit between the environment and the performer’s adaptations. As mentioned before, environmental properties may directly inform what an individual can and cannot do (Withagen et al., 2012). For example, Fink, Foo, and Warren (2009) demonstrated (by manipulating the trajectories of fly balls in a virtual environment) that how a performer gets to the right place at the right time to catch the ball is solved by relying on a strategy of cancelling optical acceleration (of the image of the ball on the catcher’s eyes). The strategy of moving in order to cancel the ball’s optical acceleration exemplifies how each player’s change in movement can be defined intrinsically by the player’s relationship with the environment (Harrison, Turvey, & Frank, 2016). The vertical optical acceleration of an approaching object can provide time-to-collision information without the need to mentally compute either distance or speed of the object to intercept it (Michaels & Zaal, 2002). An important challenge for researchers and practitioners using artificial intelligence is to capture ecophysical variables in their work to enable understanding of how intelligent behaviour might be predicated on perception- action couplings during continuous, emergent, performer-environment interactions in sports.
To sum up, for artificial intelligence to empower human intelligence, an ecological dynamics approach offers at least three suggestions. First, the sources for big data can emphasize ecophysical variables. More than the traditional emphasis in personal characteristics (e.g., height), accumulated performances (e.g., number of assists last season), or even environmental variables (e.g., size of the birth city, number of spectators), the emphasis could be placed on how the performer interacts with the circumstances of performance as captured by ecophysical variables (e.g., Carrilho et al., 2020). Second, visual analytics is highly encouraged as a tool for interpreting the data, where its display can be developed in an interdisciplinary way, with the contribution of sport scientists and practitioners, computer and data scientists, and engineers (Couceiro et al., 2016). Visual analytics and carefully developed infographics may be well placed to consider the context, in a way that an automated, context-free, output may never offer. Finally, the computation behind the analysis of big data is clearly a process that arises from the domain of expertise of computer scientists. Sport scientists typically do not have the background to programme such algorithms. However, the joint work about how the data can be analysed may fruitfully develop variables and their treatment that may better consider relevant aspects of the sport under consideration (e.g., Arau jo & Davids, 2016; Couceiro et al., 2016). Moreover, given the expertise of sport scientists to understand how movement, body, and context contributes to understand intelligent performance behaviours, they can work together with computer scientists in developing less mentalist, and more embodied and embedded, architectures for capturing the impact of artificial intelligence. This is the purpose of this book.