How should biological behaviour be modelled? A relatively new approach is to investigate problems in neuroethology by building physical robot models of biological sensorimotor systems. The explication and justification of this approach are here placed within a framework for describing and comparing models in the behavioural and biological sciences. First, simulation models – the representation of a hypothesis about a target system – are distinguished from several other relationships also termed “modelling” in discussions of scientific explanation. Seven dimensions on which simulation models can differ are defined and distinctions between them discussed:

1. Relevance: whether the model tests and generates hypotheses applicable to biology.

2. Level: the elemental units of the model in the hierarchy from atoms to societies.

3. Generality: the range of biological systems the model can represent.

4. Abstraction: the complexity, relative to the target, or amount of detail included in the model.

5. Structural accuracy: how well the model represents the actual mechanisms underlying the behaviour.

6. Performance match: to what extent the model behaviour matches the target behaviour.

7. Medium: the physical basis by which the model is implemented.

No specific position in the space of models thus defined is the only correct one, but a good modelling methodology should be explicit about its position and the justification for that position. It is argued that in building robot models biological relevance is more effective than loose biological inspiration; multiple levels can be integrated; that generality cannot be assumed but might emerge from studying specific instances; abstraction is better done by simplification than idealisation; accuracy can be approached through iterations of complete systems; that the model should be able to match and predict target behaviour; and that a physical medium can have significant advantages. These arguments reflect the view that biological behaviour needs to be studied and modelled in context, that is, in terms of the real problems faced by real animals in real environments.

It is proposed here that Webb's ideas about robots as possible models of animals need some rethinking. In our view, even though widely used biorobotics strategies are fairly successful at reproducing the macroscopic behavior of biological systems, there are still several problems unresolved on the side of robotics as well as biology. Both mathematical and hardware-like robotics models should be feasible physiologically. Control principles elaborated in robotics are not necessarily applied to biological control systems. Although observations of flying birds inspired aerodynamics and thus modern airplanes, little knowledge has been added to the neurophysiological principles underlying flight in birds. Chess playing computers might outperform most chess players, but they cannot be considered as physiologically feasible models of human thinking.

As models of living beings acting in a real world biorobots undergo an accelerated “philogenic” complexification. The first efficient robots performed simple animal behaviours (e.g., those of ants, crickets) and later on isolated elementary behaviours of complex beings. The increasing complexity of the tasks robots are dedicated to is matched by an increasing complexity and versatility of the architectures now supporting conditioning or even elementary planning.

Using the example of the difficulties which emerge when trying to model complex behaviors – such as emotional expression – that result from stochastic interactions between different components, we argue that biorobotics may well describe one possible evolution of certain features of a biological system, but cannot pretend to be a simulation of the whole behavior of the system.

Much confusion exists within the robotics and the biology communities regarding the definition of biorobotics and the aims and strategies that characterize this approach. Not even the basic criteria for identifying biorobotic research are being applied consistently. Barbara Webb has taken a crucial step towards setting a common ground from which biorobotic systems can be described, analyzed, and compared.

The question “Are ‘biorobots' good models of biological behaviour?” can be seen as a specific instance of a more general question about the relation between computer programs and models, between models and theories, and between theories and reality. This commentary develops a personal view of these relations, from an antirealism perspective. Programs, models, theories and reality are separate and distinct entities which may converge in particular cases but should never be confused.

Although the advantage of biorobotics over traditional modelling tools is not always evident from the studies on animal search addressed in the target article, this commentary argues that testing different robotic architectures and specific biological organisms in structured search spaces, where environmental constraints matter, might prove one of the most promising research strategies in comparative cognition.

The idea that biorobots can be used as a testbed for the evaluation of hypotheses about how an animal functions is supported. Generation of realistic feedback is a major advantage of biorobotic models. Nevertheless, skeptics can only be convinced that this approach is valid if significant biological insights are generated from its application.

After identifying similarities in the paradigmatic problems of biorobotics and ecological psychology, we suggest a way to compare the performance of robots with that of their biological targets. The crucial comparison is between the intentional dynamics of the robot and those of the targeted animal, a measure that depends critically on recognizing and describing the underlying affordance-effectivity match of the target system.

Robots, as well as software agents, can be of use in biology as implementations of a theory rather than as simulations of specific real world target systems. Such implementations generate hypotheses rather than representing them. Their behavior is not predicted, but rather observed, and is not expected to duplicate that of a target system. Scientific knowledge is gained through the testing of generated hypotheses.

A slight modification of Webb's diagrammatic representation of the dimensions for describing models is proposed which extends it to cover a range of theoretical models as well as material models. It is also argued that beyond a certain level robotic simulations could offer a number of real advantages over computer simulations of organisms interacting with their environment.

There is no best scientific model of anything; there are only models more or less good for different purposes. Thus, there is no general answer to the question of whether one should model biological behavior using computer simulations or robots. It all depends on what one wants to learn. This is not a question about models, but about scientific goals.

No. Animals' primary problem is the shaping of movements, guided by and adapting to sensory signals. This requires a narrower class of biorobotic models than that spanned by Webb's dimensions and examples. We claim that all model variables and mechanisms must have real counterparts, input vectors must model known sensor fields, internal state vectors and transformations must model neurophysiological processes, and output vectors must model coordinated muscle signals.

This commentary addresses an additional feature of biorobotic modeling: usefulness in prosthetic design. By implementing structurally and behaviorally accurate models as prosthetics, biologically accurate restoration can be approximated. This has potential to reestablish important peripheral elements of the sensorimotor control system, including limb biomechanics, proprioception, and vision. Examples given include musculoskeletal prosthetics and retinal implants.

As modelling becomes a popular approach in the study of biological systems it is necessary to clarify its concepts and dimensions. This helps to characterize and to distinguish models but cannot establish their quality. The virtue of a model depends on the insight gained in respect to a specific scientific question, and it is hard to measure this with a theory.

Aristotle's four causes frame Webb's question. Comprehension requires specification of trigger, function, mechanism, and representation. Robots are real models of function. Physical, biological, and epigenetic constraints delimit the hypothesis space for candidate mechanisms. Robots constitute a simplified system more susceptible to formal representation than the target system. They thus constitute an important tool in a constructivist development of scientific knowledge.

Biorobots may model the causation of relatively simple behaviors, but many animal behaviorists are concerned with complex cognitive traits and their evolution. Biorobotics seems limited in its ability to model cognition and to provide evolutionary explanations. Also, if robots could model complex traits, such as theory of mind, underdetermination could be problematic. Underdetermination is also a challenge for comparative psychologists.

Models of biological and political systems share in common an irreconcilable tension. They must extract the essential features of the system to make its workings comprehensible to the human observer, and yet the omitted underbrush is essential to the workings of the system. A good model accommodates both the workings of the system and the cognitive makeup of the observer.

Animals detect and acquire resources through a sequence of shape changes. This process is tightly coupled to the sensory and mechanical ecology of the animal. Building physical models allow us to prescind from modeling these aspects of the environment, which may not yet be described or suitably abstracted. The significance of this hybrid of physical modeling and experimentation to the acquisition of scientific knowledge is discussed.

Although robots can contribute to the understanding of biological behavior, they fail to model the processes by which humans cope with their environment. Both development and learning are characterized by complex relationships that require constant modification. Given present technology, robots can only model behaviors in specific situations and during discrete stages. Robots cannot master the complex relationships that are the hallmark of human behavior.