Elsevier

Cognition

Volume 181, December 2018, Pages 58-64
Cognition

Original Articles
Can infants' sense of agency be found in their behavior? Insights from babybot simulations of the mobile-paradigm

https://doi.org/10.1016/j.cognition.2018.07.006Get rights and content

Abstract

The development of a sense of agency is essential for understanding the causal structure of the world. Previous studies have shown that infants tend to increase the frequency of an action when it is followed by an effect. This was shown, for instance, in the mobile-paradigm, in which infants were in control of moving an overhead mobile by means of a ribbon attached to one of their limbs. These findings have been interpreted as evidence for a sense of agency early in life, as infants were thought to have detected the causal action-movement relation. We argue that solely the increase in action frequency is insufficient as evidence for this claim. Computer simulations are used to demonstrate that systematic, limb-specific increase in movement frequency found in mobile-paradigm studies can be produced by an artificial agent (a ‘babybot’) implemented with a mechanism that does not represent cause-effect relations at all. Given that a sense of agency requires representing one's actions as the cause of the effect, a behavior that is reproduced with this non-representational babybot can be argued to be, in itself, insufficient as evidence for a sense of agency. However, a behavioral pattern that to date has received little attention in the context of sense of agency, namely an additional increase in movement frequency after the action-effect relation is discontinued, is not produced by the babybot. Future research could benefit from focusing on patterns whose production cannot be reproduced by our babybot as these may require the capacity for causal learning.

Introduction

The development of a sense of agency, the feeling that one’s actions cause events, enables infants to learn from interacting with the world in ways that would not be possible otherwise. Developing a sense of agency allows for differentiating one’s own efficacy from someone else’s, which is an important step for social-cognitive skills such as imitation, theory of mind, and perspective-taking (Burrows et al., 2016, David et al., 2008, de Guzman et al., 2016, Jeannerod, 2003, Rochat and Striano, 2000). Moreover, learning about one’s own efficacy, and actively using this efficacy to intervene and discover causal structures in the world benefits causal learning (e.g., Lagnado and Sloman, 2002, Pearl, 2000, Schulz et al., 2007, Steyvers et al., 2003). Once infants have developed a sense of agency and know that their actions can cause effects, they can change and terminate their behavior to actively explore whether they cause specific effects through intervention. Since developmental science often obtains meaningful information from behavioral measures rather than for example neural measures or verbal report, developmental scientists have attempted to infer the existence of a sense of agency in young infants based on behavioral patterns. In this paper, we argue that the sense of agency cannot be inferred from the behavioral patterns previously taken as evidence. To elucidate our theoretical argument, we supported our line of reasoning with a simple computer simulation that serves as a proof of concept.

Previous studies on the developing sense of agency have shown that 2- to 4-month-old infants increase the frequency of actions that yield an interesting or pleasant effect (e.g., Rochat and Striano, 1999, Rochat and Striano, 2000). One experimental paradigm in which the behavioral pattern of increasing the frequency of actions that are followed by an effect has been observed is the classic mobile-paradigm (Rovee & Rovee, 1969), in which infants, typically around 3 months of age, lie in a crib with a mobile hanging above their head. In the experimental phase, a ribbon connects the mobile to one of the infant’s limbs, and movement of this limb causes the mobile to move (‘connect phase’). The infants’ movement behavior in this phase is then compared to their baseline movement frequency before the ribbon was attached (‘baseline phase’; e.g., Heathcock et al., 2004, Heathcock et al., 2005, Rovee-Collier et al., 1978, Watanabe and Taga, 2006, Watanabe and Taga, 2011), to their behavior after switching the ribbon to another limb (e.g., Rovee-Collier et al., 1978, Watanabe and Taga, 2009), or to their behavior following reconnection of the ribbon after a delay (e.g., Heathcock et al., 2005, Rovee-Collier et al., 1978, Rovee-Collier et al., 1980, Watanabe and Taga, 2006). Although this paradigm is predominantly known for the insights it provided into infants’ memory capacities (e.g., Cuevas et al., 2016, Hayne and Findlay, 1995, Rovee-Collier et al., 1978, Rovee-Collier et al., 1980), it has been used in various research domains, including causal learning and the developing sense of agency in young infants (Kelso, 2016, Watanabe and Taga, 2006, Watanabe and Taga, 2009, Watanabe and Taga, 2011, Watson, 1972, Watson, 1981).

Watson, for instance, reported that 2-month-old infants increased their movement frequency when their leg was connected to the mobile by the ribbon (Watson, 1972) and interpreted this behavior as a sign that infants experienced causal control over the mobile movement (Watson, 1981). More recently, these findings were replicated and extended by Watanabe and Taga, 2006, Watanabe and Taga, 2009, Watanabe and Taga, 2011. Infants again increased their movement frequency when connected to the mobile, but while the 2-month-old infants showed more movement for all limbs when the mobile was connected, the increased frequency of movements became increasingly specific to the connected limb for 3- and 4-month-old infants (Watanabe & Taga, 2006), excluding the possibility of a general arousal effect. Watanabe and Taga (2011) interpreted their findings as evidence that infants learned the causal relation between the self-produced movements and their effects on the environment, and that the change in the infants’ behavior may be a result of a sense of self-agency. Recently, Kelso (2016) supported this interpretation, stating that when infants’ spontaneous movements lead to mobile motion, movement frequency increases as they discover themselves as causal agents. Based on these behavioral patterns, researchers have thus concluded that infants already in the first few months of life experience a sense of agency.

Although infants’ behavioral patterns may indicate that they have learned the causal relation between movements and their effects, it is not self-evident that this behavior is necessarily the product of an underlying causal model. While from the perspective of the researcher we objectively know that the infants indeed cause the effect to happen, these young infants may not actually share this representation of the situation. In the context of learning new actions, for instance, Kenward (2010) argued that we should not exclude the possibility of reinforcement learning, leading to actions to be carried out habitually without a representation of the action-effect relation. Since the underlying mechanism of this behavior is unobservable and infants’ cognitive capacities are not yet as developed as they are in adults, the interpretation of infants’ behavior is not straightforward. Rather than only considering what is happening objectively, the interpretation needs to also take into account whether a displayed behavior necessitates the required representations and processes for the postulated cognitive capacity. Such an approach prevents researchers from over-interpreting infants’ behavior and provides a more nuanced view on infants’ developmental trajectory by acknowledging what representations infants can build at a given moment in their development.

Here, we address this issue in the context of the sense of agency. We evaluate whether an increased action frequency in the connect phase necessarily implies that infants have built a causal representation of their action and its effects. Putting forth the behavioral data patterns found in the mobile-paradigm as proof for a sense of agency would mean that these data patterns must exclusively be produced by a mechanism that has the capacity to generate the requisite causal representation, and not by any other cognitively plausible mechanism. To investigate what underlying mechanisms are capable of producing certain behaviors, computer simulations serve as a useful tool; in a simulation the mechanisms are precisely specified and can easily be manipulated. In this particular case, we used a computer simulation to demonstrate that the increase in limb-specific movement frequency found in mobile-paradigm studies can be accounted for by a non-representational mechanism – a mechanism that cannot internally represent cause-effect relations at all – and consequently cannot serve as evidence for a sense of agency. This simulation is not built as a computational model of infant cognition, but serves as a tool to demonstrate whether the behavior can, in principle, be generated by a more parsimonious or even an insufficient mechanism. To that end, we implemented a non-representational learning mechanism in our simulated infant, henceforth ‘babybot’. As we made the babybot unable to learn or represent cause-effect relations, any infant behavior that it reproduces is insufficient to be taken as evidence for the presence of causal representations, and hence a sense of agency. While the simulated learning mechanism might explain infants’ behavior, this simulation does not aim to test whether this is the or the only mechanism at work during the mobile-paradigm experiments. Rather, our goal was to challenge the causal interpretation of the observed behavior in infants. We thus assess the claim that the movement increase in the connect phase can only be generated by a causal learning mechanism.

We compared the babybot’s behavior to the behavior of infants in the mobile-paradigm (see Fig. 1). We focused specifically on the reproducibility of two patterns of behavior that have previously been interpreted as evidence for having learned about one’s own efficacy: (1) the increased movement frequency in the connect phase, and (2) the specificity of this increase to the connected limb. We also examined the babybot’s behavior after disconnecting the ribbon. In mobile-paradigm experiments with infants, an additional increase of movement frequency is sometimes observed in this disconnect phase (e.g., Alessandri et al., 1990, Heathcock et al., 2004, Rovee-Collier et al., 1978). The presence of this behavior as well as its potential interpretation has been ignored in previous literature on the developing sense of agency. In fact, the disconnect phase has occasionally even been omitted from the paradigm when used and interpreted in the context of a sense of agency (e.g., Watanabe & Taga, 2011), although we believe it may contain useful information regarding infants’ representation of the situation as it could indicate that infants are trying to reproduce the effect and, in fact, did learn the causal relation.

Section snippets

Babybot

We simulated a simplified version of the mobile-paradigm programmed with MATLAB (see supplementary materials for pseudocode and https://github.com/LZaadnoordijk/Babybot for the MATLAB code). For simplicity, we limited the babybot’s movement possibilities for each of its four independent limbs to three motor signals: move up, move down and hold still. Furthermore, the babybot had three possible positions for each limb: top, middle, bottom. By selecting a motor signal, the babybot could change

Results

The simulation results show that the babybot displays two of the behavioral patterns of interest (see Fig. 3). It moved the connected limb more frequently than both the limb’s own baseline and the other unconnected limbs. This demonstrates that the infant behavior during the connect phase can be produced by the non-representational learning mechanism implemented in the babybot. These results thus show that a non-representational learning mechanism can account for the infant behavior that has

Discussion

The development of a sense of agency is an important milestone in the early months of life because of its role in social development and causal learning. Previous studies on the developing sense of agency have shown that infants as young as two months of age increase the frequency of an action when this action is followed by an effect (e.g., Gergely and Watson, 1999, Rochat and Striano, 1999, Watanabe and Taga, 2006, Watanabe and Taga, 2009, Watanabe and Taga, 2011, Watson, 1972). These

Declarations of interest

None.

Acknowledgments

The authors thank Tobias Winner and Dennis Merkus for programming the simulation platform that formed the basis of our simulations. We further thank Iris van Rooij, Tarek R. Besold and Emma Ward for their valuable contributions to this manuscript.

References (35)

  • H. Watanabe et al.

    General to specific development of movement patterns and memory for contingency between actions and events in young infants

    Infant Behavior and Development

    (2006)
  • H. Watanabe et al.

    Flexibility in infant actions during arm-and leg-based learning in a mobile paradigm

    Infant Behavior and Development

    (2009)
  • H. Watanabe et al.

    Initial-state dependency of learning in young infants

    Human Movement Science

    (2011)
  • S.M. Alessandri et al.

    Violation of expectancy and frustration in early infancy

    Developmental Psychology

    (1990)
  • C.A. Burrows et al.

    Functional connectivity of brain regions for self- and other-evaluation in children, adolescents and adults with autism

    Developmental Science

    (2016)
  • K. Cuevas et al.

    A dissociation between recognition and reactivation: The renewal effect at 3 months of age

    Developmental Psychobiology

    (2016)
  • M. de Guzman et al.

    Self–other control processes in social cognition: From imitation to empathy

    Philosophical Transactions of the Royal Society B

    (2016)
  • Cited by (21)

    • Correlation between performance and quantity/variability of leg exploration in a contingency learning task during infancy

      2023, Infant Behavior and Development
      Citation Excerpt :

      Future studies should include more infants to elucidate the relationship between each variable and the learning outcome. Lastly, it is an important and tricky question to understand whether infants intentionally adapted their behavior to perform the task or if it was merely an automated response pattern (Zaadnoordijk et al., 2018). Neuroimaging methods may help measure brain function during the learning process for future studies.

    • Do infants have agency? – The importance of control for the study of early agency

      2022, Developmental Review
      Citation Excerpt :

      It was observed in a few studies that in this disconnect phase, the leg movement increased rapidly and then dropped to the frequency level observed in the baseline condition (Lewis et al., 1990; Heathcock et al., 2004; Rovee-Collier et al., 1978; Zaadnoordijk et al., 2018, 2020), an effect that has been referred to as the extinction burst. This effect has only been observed in a handful of studies (see Table 3) and has only recently been investigated more systematically (Zaadnoordijk et al., 2018, 2020) (see Table 3). Immediately after being freed from the connection to the mobile, infants’ overall limb movement increases abruptly.

    • From movement to action: An EEG study into the emerging sense of agency in early infancy

      2020, Developmental Cognitive Neuroscience
      Citation Excerpt :

      This behavior was replicated in the current study, as infants showed a linear increase in movement frequency in the connect phase. Computer simulation research, however, has demonstrated that an underlying causal model cannot be inferred from this behavioral pattern (Zaadnoordijk et al., 2018). This was why, in contrast to previous research, our focus was on the violation of expectation, as this indicates that infants have made a prediction regarding the consequences of their action and thus have built an internal model.

    View all citing articles on Scopus
    View full text