In the target article, Kline has done a laudable job putting forward a comprehensive framework for discussing teaching from an evolutionary perspective. Her taxonomy of adaptive problems that arise in social learning is defined by two factors: insufficient attention and/or access on the part of the pupil. Using this framework, Kline identifies “active direct teaching” – which she introduces as akin to natural pedagogy (Csibra & Gergely Reference Csibra and Gergely2011) – as an adaptation evolved in response to the problem of transferring knowledge to a pupil who “lacks both attention and access to a relevant stimulus or information” (sect. 3.1.5, para. 1).
Kline makes a convincing case for how shortages of attention/access to stimuli can be dealt with by the teaching mechanisms whose function it is to establish sufficient conditions for individual social-observational learning: social tolerance, opportunity provisioning, and stimulus/local enhancement. But what Kline means by the lack of access/attention to information is less clear. We appreciate how organizing the taxonomy along simple lines helps in presenting an integrative framework. However, we worry that in this case the simplicity comes at the cost of missing on the opportunity to specify the adaptive problems in response to which teaching might have evolved in humans, and to recognize how they relate to what humans teach about.
There are at least four types of adaptive learnability problems that have been identified in human social learning and hypothesized to drive, in humans, the emergence of a system of adaptations, which facilitates teaching and learning by means of communication (Csibra & Gergely Reference Csibra and Gergely2011). All four arise from the inherent opacity of human cultural knowledge. First, observed behavior of the model has to be interpreted by the learner as a goal-directed action with a recognized goal. In a complex cultural environment, a learner may thus often face the problem of teleological opacity. For example, if the action-sequence involves many steps with individual sub-goals that provide no immediate benefits to the model and the learner (e.g., when preparation of food requires pre-processing the ingredients), its overarching goal may be difficult to grasp. Second, the learner may often face a challenge of interpreting an action despite its causal opacity. This issue is pertinent, for example, to learning manners of tool-use or ritual behaviors, whose causal relations to the stipulated goal may be opaque both to the learner and the model.
Third, although any observation is by necessity bounded to the observed particulars (e.g., a particular tool-item), the learner faces uncertainty as to whether the acquired information (e.g., a function inferred from the outcome of the observed action with the tool) should be treated as applying exclusively to the observed individual items (e.g., it is a dispositional property of this tool to bringing about this outcome), or whether it should be treated as generic information about the kind manifested by the particular and, therefore, generalizable to other tokens of this kind (e.g., the dispositional property to bring about this type of outcome is the function of this kind of tool). Fourth, in addition to the uncertainty about genericity of the acquired information, the learner may also be uncertain about its sharedness, that is whether, what she observed (e.g., a means action or an emotion expressed towards something) is to be attributed solely to the model or to should be treated as knowledge of a cultural practice shared by other group-members. Notably, all of these learnability problems arise because abstract generic representations with opaque contents, such as tool-kinds and social norms, can become the content of human social learning.
Admittedly, a pupil could move towards overcoming these learnability problems through a laborious process of inductive inferences based on individual social-observational learning guided by stimulus/local enhancements. During a demonstration, the model often shows and points at things for the pupil, and these behaviors may enhance the pupil's motivation and attention to stimuli. Notably, this particular function of communicative signals was singled out in Kline's examples and her depiction of natural pedagogy as a “motivational system” (sect. 2.3, para. 4).
However, numerous recent empirical results suggest an additional, different role of such ostensive-communicative signals (i.e. behaviors, from which the pupil can infer that she is the addressee of the model's communication) in human social learning. In human children, demonstrations that involve ostensive signals promote (i) acquiring causally opaque means actions (e.g., turning on the lamp with the forehead rather than the hands; Király et al. Reference Király, Csibra and Gergely2013); (ii) encoding kind-relevant permanent features of objects (e.g., shape) at the expense of transient episodic properties (e.g., current location or relative numerosity), that do not contribute relevant information to kind membership (Chen et al. Reference Chen, Volein, Gergely and Csibra2011; Yoon et al. Reference Yoon, Johnson and Csibra2008); (iii) forming generic representations resilient to counterevidence (e.g., encoding a tool's function as enduring, even if the tool currently doesn't work or was temporarily put to a different use; Butler & Markman Reference Butler and Markman2012; Hernik & Csibra Reference Hernik and Csibra2015); and (iv) interpreting object-directed emotional expressions as evidence of shared knowledge applicable to interactions with other individuals (e.g., “this object is not desirable,” rather than “she doesn't like it”; Egyed et al.. Reference Egyed, Király and Gergely2013; Träuble & Bätz Reference Träuble and Bätz2014). According to Gergely and Jacob (2012), ostensive communication may also “fast-track” learning about kinds from statistically meager evidence by inviting in the learner an assumption of “strong,” non-random sampling on the part of the demonstrator (Gweon et al. Reference Gweon, Tenenbaum and Schulz2010).
The empirical results suggest that rather than merely shifting attention- and motivation-related parameters of individual observational learning, teaching through ostensive communication changes how the learner interprets the stimulus. Ostensive and referential signals license the pupil to engage in relevance-guided inferences, which afford an interpretation that goes beyond the particular stimuli of the demonstration episode. In communicative demonstrations, humans use the concrete and particular to teach about the abstract and generic.
We applaud Kline's endeavor of characterizing forms of teaching through their adaptive functions. However, a comprehensive account of the adaptive problems behind teaching adaptations should recognize their relation to the types of transmitted contents. Adaptive problems in knowledge transfer arise not only because human children may lack attention or access, but critically because they often have to infer opaque contents as shared generic knowledge.
In the target article, Kline has done a laudable job putting forward a comprehensive framework for discussing teaching from an evolutionary perspective. Her taxonomy of adaptive problems that arise in social learning is defined by two factors: insufficient attention and/or access on the part of the pupil. Using this framework, Kline identifies “active direct teaching” – which she introduces as akin to natural pedagogy (Csibra & Gergely Reference Csibra and Gergely2011) – as an adaptation evolved in response to the problem of transferring knowledge to a pupil who “lacks both attention and access to a relevant stimulus or information” (sect. 3.1.5, para. 1).
Kline makes a convincing case for how shortages of attention/access to stimuli can be dealt with by the teaching mechanisms whose function it is to establish sufficient conditions for individual social-observational learning: social tolerance, opportunity provisioning, and stimulus/local enhancement. But what Kline means by the lack of access/attention to information is less clear. We appreciate how organizing the taxonomy along simple lines helps in presenting an integrative framework. However, we worry that in this case the simplicity comes at the cost of missing on the opportunity to specify the adaptive problems in response to which teaching might have evolved in humans, and to recognize how they relate to what humans teach about.
There are at least four types of adaptive learnability problems that have been identified in human social learning and hypothesized to drive, in humans, the emergence of a system of adaptations, which facilitates teaching and learning by means of communication (Csibra & Gergely Reference Csibra and Gergely2011). All four arise from the inherent opacity of human cultural knowledge. First, observed behavior of the model has to be interpreted by the learner as a goal-directed action with a recognized goal. In a complex cultural environment, a learner may thus often face the problem of teleological opacity. For example, if the action-sequence involves many steps with individual sub-goals that provide no immediate benefits to the model and the learner (e.g., when preparation of food requires pre-processing the ingredients), its overarching goal may be difficult to grasp. Second, the learner may often face a challenge of interpreting an action despite its causal opacity. This issue is pertinent, for example, to learning manners of tool-use or ritual behaviors, whose causal relations to the stipulated goal may be opaque both to the learner and the model.
Third, although any observation is by necessity bounded to the observed particulars (e.g., a particular tool-item), the learner faces uncertainty as to whether the acquired information (e.g., a function inferred from the outcome of the observed action with the tool) should be treated as applying exclusively to the observed individual items (e.g., it is a dispositional property of this tool to bringing about this outcome), or whether it should be treated as generic information about the kind manifested by the particular and, therefore, generalizable to other tokens of this kind (e.g., the dispositional property to bring about this type of outcome is the function of this kind of tool). Fourth, in addition to the uncertainty about genericity of the acquired information, the learner may also be uncertain about its sharedness, that is whether, what she observed (e.g., a means action or an emotion expressed towards something) is to be attributed solely to the model or to should be treated as knowledge of a cultural practice shared by other group-members. Notably, all of these learnability problems arise because abstract generic representations with opaque contents, such as tool-kinds and social norms, can become the content of human social learning.
Admittedly, a pupil could move towards overcoming these learnability problems through a laborious process of inductive inferences based on individual social-observational learning guided by stimulus/local enhancements. During a demonstration, the model often shows and points at things for the pupil, and these behaviors may enhance the pupil's motivation and attention to stimuli. Notably, this particular function of communicative signals was singled out in Kline's examples and her depiction of natural pedagogy as a “motivational system” (sect. 2.3, para. 4).
However, numerous recent empirical results suggest an additional, different role of such ostensive-communicative signals (i.e. behaviors, from which the pupil can infer that she is the addressee of the model's communication) in human social learning. In human children, demonstrations that involve ostensive signals promote (i) acquiring causally opaque means actions (e.g., turning on the lamp with the forehead rather than the hands; Király et al. Reference Király, Csibra and Gergely2013); (ii) encoding kind-relevant permanent features of objects (e.g., shape) at the expense of transient episodic properties (e.g., current location or relative numerosity), that do not contribute relevant information to kind membership (Chen et al. Reference Chen, Volein, Gergely and Csibra2011; Yoon et al. Reference Yoon, Johnson and Csibra2008); (iii) forming generic representations resilient to counterevidence (e.g., encoding a tool's function as enduring, even if the tool currently doesn't work or was temporarily put to a different use; Butler & Markman Reference Butler and Markman2012; Hernik & Csibra Reference Hernik and Csibra2015); and (iv) interpreting object-directed emotional expressions as evidence of shared knowledge applicable to interactions with other individuals (e.g., “this object is not desirable,” rather than “she doesn't like it”; Egyed et al.. Reference Egyed, Király and Gergely2013; Träuble & Bätz Reference Träuble and Bätz2014). According to Gergely and Jacob (2012), ostensive communication may also “fast-track” learning about kinds from statistically meager evidence by inviting in the learner an assumption of “strong,” non-random sampling on the part of the demonstrator (Gweon et al. Reference Gweon, Tenenbaum and Schulz2010).
The empirical results suggest that rather than merely shifting attention- and motivation-related parameters of individual observational learning, teaching through ostensive communication changes how the learner interprets the stimulus. Ostensive and referential signals license the pupil to engage in relevance-guided inferences, which afford an interpretation that goes beyond the particular stimuli of the demonstration episode. In communicative demonstrations, humans use the concrete and particular to teach about the abstract and generic.
We applaud Kline's endeavor of characterizing forms of teaching through their adaptive functions. However, a comprehensive account of the adaptive problems behind teaching adaptations should recognize their relation to the types of transmitted contents. Adaptive problems in knowledge transfer arise not only because human children may lack attention or access, but critically because they often have to infer opaque contents as shared generic knowledge.
ACKNOWLEDGMENT
This work is supported by an Advanced Grant (OSTREFCOM) from the European Research Council and by a grant from the Hungarian Science Foundation (OTKA Grant #: NK 83997).