Biological and artificial cognition: what can we learn about mechanisms by modelling physical cognition problems using artificial intelligence planning techniques?

Do we fully understand the structure of the problems we present to our subjects in experiments on animal cognition, and the information required to solve them? While we currently have a good understanding of the behavioural and neurobiological mechanisms underlying associative learning processes, we understand much less about the mechanisms underlying more complex forms of cognition in animals. In this study, we present a proposal for a new way of thinking about animal cognition experiments. We describe a process in which a physical cognition task domain can be decomposed into its component parts, and models constructed to represent both the causal events of the domain and the information available to the agent. We then implement a simple set of models, using the planning language MAPL within the MAPSIM simulation environment, and applying it to a puzzle tube task previously presented to orangutans. We discuss the results of the models and compare them with the results from the experiments with orangutans, describing the advantages of this approach, and the ways in which it could be extended.     

Morgan's Canon: Is it still a useful rule of thumb?

For over 100 years, Morgan's Canon has served as the criterion for distinguishing what appears to be complex cognitive processes shown by animals from simpler associative learning processes (Pavlovian and instrumental conditioning). Morgan's ( 1894 ) canon states “In no case is an animal activity to be interpreted in terms of higher psychological processes if it can be fairly interpreted in terms of processes which stand lower in the scale of psychological evolution and development.” In the present article, several examples are provided in which complex human‐like processes are proposed to have been demonstrated but the judicious use of Morgan's Canon suggests that simpler mechanisms may be sufficient to account for the behavior. The use of Morgan's Canon is not meant to reduce behavior to its lowest common denominator but rather to challenge investigators to develop procedures that can distinguish between simple behavioral principles and the more complex processes that cannot be explained in terms of genetics or simple conditioning. Whatever the results of these experiments, they should help identify the underlying processes and mechanisms involved.     

Towards neuro-inspired symbolic models of cognition: linking neural dynamics to behaviors through asynchronous communications

A computational architecture modeling the relation between perception and action is proposed. Basic brain processes representing synaptic plasticity are first abstracted through asynchronous communication protocols and implemented as virtual microcircuits. These are used in turn to build mesoscale circuits embodying parallel cognitive processes. Encoding these circuits into symbolic expressions gives finally rise to neuro-inspired programs that are compiled into pseudo-code to be interpreted by a virtual machine. Quantitative evaluation measures are given by the modification of synapse weights over time. This approach is illustrated by models of simple forms of behaviors exhibiting cognition up to the third level of animal awareness. As a potential benefit, symbolic models of emergent psychological mechanisms could lead to the discovery of the learning processes involved in the development of cognition. The executable specifications of an experimental platform allowing for the reproduction of simulated experiments are given in “Appendix”.     

The evolution of associative learning: A factor in the Cambrian explosion

The Cambrian explosion is probably the most spectacular diversification in evolutionary history, and understanding it has been a challenge for biologists since the time of Darwin. We propose that one of the key factors that drove this great diversification was associative learning. Although the evolutionary emergence of associative learning required only small modifications in already existing memory mechanisms and may have occurred in parallel in several groups, once this type of learning appeared on the evolutionary scene, it led to extreme diversifying selection at the ecological level: it enabled animals to exploit new niches, promoted new types of relations and arms races, and led to adaptive responses that became fixed through genetic accommodation processes. This learning-based diversification was accompanied by neurohormonal stress, which led to an ongoing destabilization and re-patterning of the epigenome, which, in turn, enabled further morphological, physiological, and behavioral diversification. Our hypothesis combines several previous ideas about the dynamics of the Cambrian explosion and provides a unifying framework that includes both ecological and genomic factors. We conclude by suggesting research directions that would clarify the timing and manner in which associative learning evolved, and the effects it had on the evolution of nervous systems, genomes, and animal morphology.     

Age-Related Decline in Associative Learning in Healthy Chinese Adults

Paired associates learning (PAL) has been widely used in aging-related research, suggesting an age-related decline in associative learning. However, there are several cognitive processes (attention, spatial and recognition memory, strategy, and associative learning) involved in PAL. It is unclear which component contributes to the decline in PAL performance associated with age effects. The present study determines whether age effects on associative learning are independent of other cognitive processes involved in PAL. Using a validated computerized cognitive program (CANTAB), we examined cognitive performance of associative learning, spatial and recognition memory, attention and strategy use in 184 Singaporean Chinese adults aged from 21 to 80 years old. Linear regression revealed significant age-related decline in associative learning, spatial and recognition memory, and the level of strategy use. This age-related decline in associative learning remains even after adjusting for attention, spatial and recognition memory, and strategy use. These results show that age effects on associative learning are independent of other cognitive processes involved in PAL.     

Eurasian jays (Garrulus glandarius) overcome their current desires to anticipate two distinct future needs and plan for them appropriately

Western scrub-jays (Aphelocoma californica) have been shown to overcome present satiety to cache food they will desire in the future. Here, we show that another corvid, the Eurasian jay (Garrulus glandarius), can distinguish between two distinct future desires and plan for each appropriately, despite experiencing a conflicting current motivation. We argue that these data address the criticisms of previous work, and suggest a way in which associative learning processes and future-oriented cognition may combine to allow prospective behaviour.     

Are cognition and personality related in budgerigars?

In recent years, researchers have been attempting to relate differences in personality (e.g., boldness, aggressiveness, exploration tendency) to variation in cognition (performances in tasks that require learning, reasoning, attention, or memory, etc.) both theoretically and empirically. However, it is unclear on what basis personality and cognition might be associated with each other. Previous theory suggests a connection between fast–slow personality types and cognitive speed–accuracy tradeoffs. In this study, we tested this hypothesis in budgerigars and found that, in their 1st associative learning, birds with fast personality (less fearful of handling stress) were fast learners in the beginning, while slow personality individuals improved faster, but both types of birds did not differ in accuracy. However, these relationships were context-dependent. No significant relationship was found in subsequent learning tasks (reversal learning and a 2nd associative learning) in the familiar context (task setup and apparatus similar to the 1st associative learning). We then conducted a problem-solving experiment with novel setup and apparatus to test 1 possible explanation that the association between personality and cognition in the 1st associative learning might be caused by noncognitive constraint, such as fearfulness when facing novel task setup and apparatus. We found that fast individuals interacted more with the problem box and solved it, whereas the slow birds were not. We suggest that personalities can influence cognitive performances and trigger a cognitive speed-improvement tradeoff under the novel context. However, there are no consistent cognitive styles that co-varied with different personalities.     

Simple minds: a qualified defence of associative learning

Using cooperation in chimpanzees as a case study, this article argues that research on animal minds needs to steer a course between 'association-blindness'-the failure to consider associative learning as a candidate explanation for complex behaviour-and 'simple-mindedness'-the assumption that associative explanations trump more cognitive hypotheses. Association-blindness is challenged by the evidence that associative learning occurs in a wide range of taxa and functional contexts, and is a major force guiding the development of complex human behaviour. Furthermore, contrary to a common view, association-blindness is not entailed by the rejection of behaviourism. Simple-mindedness is founded on Morgan's canon, a methodological principle recommending 'lower' over 'higher' explanations for animal behaviour. Studies in the history and philosophy of science show that Morgan failed to offer an adequate justification for his canon, and subsequent attempts to justify the canon using evolutionary arguments and appeals to simplicity have not been successful. The weaknesses of association-blindness and simple-mindedness imply that there are no short-cuts to finding out about animal minds. To decide between associative and yet more cognitive explanations for animal behaviour, we have to spell them out in sufficient detail to allow differential predictions, and to test these predictions through observation and experiment.     

Genetic variation in brain-derived neurotrophic factor and human fear conditioning

Brain-derived neurotrophic factor (BDNF) has been implicated in hippocampal-dependent learning processes, and carriers of the Met allele of the Val66Met BDNF genotype are characterized by reduced hippocampal structure and function. Recent nonhuman animal work suggests that BDNF is also crucial for amygdala-dependent associative learning. The present study sought to examine fear conditioning as a function of the BDNF polymorphism. Fifty-seven participants were genotyped for the BDNF polymorphism and took part in a differential-conditioning paradigm. Participants were shocked following a particular conditioned stimulus (CS+) and were also presented with stimuli that ranged in perceptual similarity to the CS+ (20, 40 or 60% smaller or larger than the CS+). The eye blink component of the startle response was measured to quantify fear conditioning; post-task shock likelihood ratings for each stimulus were also obtained. All participants reported that shock likelihood varied with perceptual similarity to the CS+ and showed potentiated startle in response to CS ± 20% stimuli. However, only the Val/Val group had potentiated startle responses to the CS+. Met allele carrying individuals were characterized by deficient fear conditioning – evidenced by an attenuated startle response to CS+ stimuli. Variation in the BDNF genotype appears related to abnormal fear conditioning, consistent with nonhuman animal work on the importance of BDNF in amygdala-dependent associative learning. The relation between genetic variation in BDNF and amygdala-dependent associative learning deficits is discussed in terms of potential mechanisms of risk for psychopathology.     

The evolution of intelligence: adaptive specializations versus general process

Darwin argued that between-species differences in intelligence were differences of degree, not of kind. The contemporary ecological approach to animal cognition argues that animals have evolved species-specific and problem-specific processes to solve problems associated with their particular ecological niches: thus different species use different processes, and within a species, different processes are used to tackle problems involving different inputs. This approach contrasts both with Darwin's view and with the general process view, according to which the same central processes of learning and memory are used across an extensive range of problems involving very different inputs. We review evidence relevant to the claim that the learning and memory performance of non-human animals varies according to the nature of the stimuli involved. We first discuss the resource distribution hypothesis, olfactory learning-set formation, and the 'biological constraints' literature, but find no convincing support from these topics for the ecological account of cognition. We then discuss the claim that the performance of birds in spatial tasks of learning and memory is superior in species that depend heavily upon stored food compared to species that either show less dependence upon stored food or do not store food. If it could be shown that storing species enjoy a superiority specifically in spatial (and not non-spatial) tasks, this would argue that spatial tasks are indeed solved using different processes from those used in non-spatial tasks. Our review of this literature does not find a consistent superiority of storing over non-storing birds in spatial tasks, and, in particular, no evidence of enhanced superiority of storing species when the task demands are increased, by, for example, increasing the number of items to be recalled or the duration of the retention period. We discuss also the observation that the hippocampus of storing birds is larger than that of non-storing birds, and find evidence contrary to the view that hippocampal enlargement is associated with enhanced spatial memory; we are, however, unable to suggest a convincing alternative explanation for hippocampal enlargement. The failure to find solid support for the ecological view supports the view that there are no qualitative differences in cognition between animal species in the processes of learning and memory. We also argue that our review supports our contention that speculation about the phylogenetic development and function of behavioural processes does not provide a solid basis for gaining insight into the nature of those processes. We end by confessing to a belief in one major qualitative difference in cognition in animals: we believe that humans alone are capable of acquiring language, and that it is this capacity that divides our intelligence so sharply from non-human intelligence.     

Reinforcement pre-exposure enhances spatial memory formation in <Emphasis Type="Italic">Drosophila</Emphasis>

Apparently unpaired exposure to appetitive or aversive stimuli can suppress or enhance later associative learning. While the suppressive effect has been found in both vertebrate and invertebrate animals, it is not clear if the enhancing effect is restricted to the vertebrates. Additionally, whether Drosophila associative learning can be influenced in either direction is open. To address these questions, we examined the effects of pre-exposing flies to a high temperature negative reinforcer in the heat-box place-learning paradigm. We found that pre-exposing flies to an unavoidable high temperature enhanced later associative conditioning that uses mild increases in temperature. This enhancement lasts at least 20 min, does not depend on changes in the straightforward avoidance behavior of a high temperature source, and is independent of the antennal thermosensor. We thus provide an example of enhanced associative learning after unpaired exposure to a typical reinforcer in an invertebrate animal, suggesting the conservation of this component of learning.     

Honeybee Associative Learning Performance and Metabolic Stress Resilience Are Positively Associated

Background

Social-environmental influences can affect animal cognition and health. Also, human socio-economic status is a covariate factor connecting psychometric test-performance (a measure of cognitive ability), educational achievement, lifetime health, and survival. The complimentary hypothesis, that mechanisms in physiology can explain some covariance between the same traits, is disputed. Possible mechanisms involve metabolic biology affecting integrity and stability of physiological systems during development and ageing. Knowledge of these relationships is incomplete, and underlying processes are challenging to reveal in people. Model animals, however, can provide insights into connections between metabolic biology and physiological stability that may aid efforts to reduce human health and longevity disparities.

Results

We document a positive correlation between a measure of associative learning performance and the metabolic stress resilience of honeybees. This relationship is independent of social factors, and may provide basic insights into how central nervous system (CNS) function and metabolic biology can be associated. Controlling for social environment, age, and learning motivation in each bee, we establish that learning in Pavlovian conditioning to an odour is positively correlated with individual survival time in hyperoxia. Hyperoxia induces oxidative metabolic damage, and provides a measure of metabolic stress resistance that is often related to overall lifespan in laboratory animals. The positive relationship between Pavlovian learning ability and stress resilience in the bee is not equally established in other model organisms so far, and contrasts with a genetic cost of improved associative learning found in Drosophila melanogaster.

Conclusions

Similarities in the performances of different animals need not reflect common functional principles. A correlation of honeybee Pavlovian learning and metabolic stress resilience, thereby, is not evidence of a shared biology that will give insight about systems integrity in people. Yet, the means to resolve difficult research questions often come from findings in distant areas of science while the model systems that turn out to be valuable are sometimes the least predictable. Our results add to recent findings indicating that honeybees can become instrumental to understanding how metabolic biology influences life outcomes.     

Stimulus coding in human associative learning: flexible representations of parts and wholes

An enduring theme for theories of associative learning is the problem of explaining how configural discriminations--ones in which the significance of combinations of cues is inconsistent with the significance of the individual cues themselves-are learned. One approach has been to assume that configurations are the basic representational form on which associative processes operate, another has tried in contrast to retain elementalism. We review evidence that human learning is representationally flexible in a way that challenges both configural and elemental theories. We describe research showing that task demands, prior experience, instructions, and stimulus properties all influence whether a particular problem is solved configurally or elementally. Lines of possible future theory development are discussed.     

The evolution of comparative cognition: Is the snark still a boojum?

In “The Snark is a Boojum”, Beach [Beach, F.A., 1950. The snark was a boojum. American Psychologist. 5, 115–124] famously asserted that animal psychology embraced too few species and too few problems to deserve the name comparative. Later in the 20th century, others [e.g. Kamil, A.C., 1988. A synthetic approach to the study of animal intelligence. In: Leger, D.W. (Ed.), Comparative Perspectives in Modern Psychology. Nebraska Symposium on Motivation, vol. 35. University of Nebraska Press, Lincoln, NE, pp. 230–257; Shettleworth, S.J., 1993. Where is the comparison in comparative cognition? Alternative research programs. Psychological Science. 4, 179–184] expressed similar concerns about the new subfield of comparative cognition, suggesting that a more biological approach to choice of species and problems was needed to balance a dominant anthropocentrism. The last 10–15 years have seen many new developments, and a recent survey like Beach’s reveals a very different picture. Not only are many more species being studied, contributions by researchers from different backgrounds are increasing, and research on comparative cognition is better connected with developmental psychology, behavioral neuroscience, primatology, behavioral ecology, and other fields. Contemporary research addresses three major aspects of cognition about equally: basic processes, physical cognition, and social cognition. This article describes a selected research program from each area, chosen to exemplify current trends and challenges for the field.     

Social cognition

Social cognition concerns the various psychological processes that enable individuals to take advantage of being part of a social group. Of major importance to social cognition are the various social signals that enable us to learn about the world. Such signals include facial expressions, such as fear and disgust, which warn us of danger, and eye gaze direction, which indicate where interesting things can be found. Such signals are particularly important in infant development. Social referencing, for example, refers to the phenomenon in which infants refer to their mothers' facial expressions to determine whether or not to approach a novel object. We can learn a great deal simply by observing others. Much of this signalling seems to happen automatically and unconsciously on the part of both the sender and the receiver. We can learn to fear a stimulus by observing the response of another, in the absence of awareness of that stimulus. By contrast, learning by instruction, rather than observation, does seem to depend upon awareness of the stimulus, since such learning does not generalize to situations where the stimulus is presented subliminally. Learning by instruction depends upon a meta-cognitive process through which both the sender and the receiver recognize that signals are intended to be signals. An example would be the 'ostensive' signals that indicate that what follows are intentional communications. Infants learn more from signals that they recognize to be instructive. I speculate that it is this ability to recognize and learn from instructions rather than mere observation which permitted that advanced ability to benefit from cultural learning that seems to be unique to the human race.     

Associative fear learning enhances sparse network coding in primary sensory cortex

Several models of associative learning predict that stimulus processing changes during association formation. How associative learning reconfigures neural circuits in primary sensory cortex to "learn" associative attributes of a stimulus remains unknown. Using 2-photon in vivo calcium imaging to measure responses of networks of neurons in primary somatosensory cortex, we discovered that associative fear learning, in which whisker stimulation is paired with foot shock, enhances sparse population coding and robustness of the conditional stimulus, yet decreases total network activity. Fewer cortical neurons responded to stimulation of the trained whisker than in controls, yet their response strength was enhanced. These responses were not observed in mice exposed to a nonassociative learning procedure. Our results define how the cortical representation of a sensory stimulus is shaped by associative fear learning. These changes are proposed to enhance efficient sensory processing after associative learning.     

Spatial inference without a cognitive map: the role of higher-order path integration

The cognitive map has been taken as the standard model for how agents infer the most efficient route to a goal location. Alternatively, path integration – maintaining a homing vector during navigation – constitutes a primitive and presumably less-flexible strategy than cognitive mapping because path integration relies primarily on vestibular stimuli and pace counting. The historical debate as to whether complex spatial navigation is ruled by associative learning or cognitive map mechanisms has been challenged by experimental difficulties in successfully neutralizing path integration. To our knowledge, there are only three studies that have succeeded in resolving this issue, all showing clear evidence of novel route taking, a behaviour outside the scope of traditional associative learning accounts. Nevertheless, there is no mechanistic explanation as to how animals perform novel route taking. We propose here a new model of spatial learning that combines path integration with higher-order associative learning, and demonstrate how it can account for novel route taking without a cognitive map, thus resolving this long-standing debate. We show how our higher-order path integration (HOPI) model can explain spatial inferences, such as novel detours and shortcuts. Our analysis suggests that a phylogenetically ancient, vector-based navigational strategy utilizing associative processes is powerful enough to support complex spatial inferences.     

Exploring action dynamics as an index of paired-associate learning

Much evidence exists supporting a richer interaction between cognition and action than commonly assumed. Such findings demonstrate that short-timescale processes, such as motor execution, may relate in systematic ways to longer-timescale cognitive processes, such as learning. We further substantiate one direction of this interaction: the flow of cognition into action systems. Two experiments explored match-to-sample paired-associate learning, in which participants learned randomized pairs of unfamiliar symbols. During the experiments, their hand movements were continuously tracked using the Nintendo Wiimote. Across learning, participant arm movements are initiated and completed more quickly, exhibit lower fluctuation, and exert more perturbation on the Wiimote during the button press. A second experiment demonstrated that action dynamics index novel learning scenarios, and not simply acclimatization to the Wiimote interface. Results support a graded and systematic covariation between cognition and action, and recommend ways in which this theoretical perspective may contribute to applied learning contexts.     

Unconventional sex: fresh approaches to courtship learning

Understanding the complex array of genes, proteins and cells involved in learning and memory is a major challenge for neuroscientists. Using the genetically powerful model system, Drosophila melanogaster, and its well-studied courtship behavior, investigators have begun to delineate essential elements of associative and nonassociative behavioral plasticity. Advances in transgenic tools and developments in behavioral assays have increased the power of studying courtship learning in the fruit fly.