Hostname: page-component-745bb68f8f-b6zl4 Total loading time: 0 Render date: 2025-02-11T16:03:38.636Z Has data issue: false hasContentIssue false

Reasoning, robots, and navigation: Dual roles for deductive and abductive reasoning

Published online by Cambridge University Press:  29 March 2011

Janet Wiles
Affiliation:
School of Information Technology & Electrical Engineering, University of Queensland, Brisbane 4072, Australia. wiles@itee.uq.edu.auhttp://www.itee.uq.edu.au/~janetw/

Abstract

Mercier & Sperber (M&S) argue for their argumentative theory in terms of communicative abilities. Insights can be gained by extending the discussion beyond human reasoning to rodent and robot navigation. The selection of arguments and conclusions that are mutually reinforcing can be cast as a form of abductive reasoning that I argue underlies the construction of cognitive maps in navigation tasks.

Type
Open Peer Commentary
Copyright
Copyright © Cambridge University Press 2011

Mercier and Sperber's (M&S's) theory of the adaptive value of argumentative reasoning is intriguing from a computational perspective, since the search for arguments that support a given conclusion is computationally more difficult (viewed as a reasoning problem) than logical reasoning. The first logical solvers were developed in the 1950s (Newell & Simon Reference Newell and Simon1956). Argumentative computers are yet to be developed.

Argumentative reasoning, defined broadly as the discovery of statements to support a given conclusion can be cast as a form of adbuctive reasoning, or inferring a precondition from a consequent (following Peirce Reference Peirce1931–35). Such reasoning is logically fallacious, but as M&S's target article details, it is typical of human behaviour to select arguments and conclusions that together are mutually reinforcing.

We accept M&S's arguments for the adaptive value of argumentative reasoning as a communicative skill. However, just as questions have been raised in other fields about the evolution of the sophisticated communicative abilities of humans, we can also ask how an argumentative ability could have evolved. Many evolutionary adaptations are thought to be exaptations; that is, new uses for existing structures. Verbal argumentative reasoning obviously draws on linguistic ability, but it need not postdate it. We consider the possibility that cognitive abilities underlying argumentative reasoning may predate the evolution of language by exapting abductive abilities from other domains.

Reasoning is not the only domain where adaptive behaviour may utilise abductive reasoning. A much more ancient evolutionary ability, which humans share with other mammals, birds, reptiles, and even insects, is the ability to navigate. Much is known about the navigational systems of mammals, including the neural representations of places (O'Keefe & Dostrovsky Reference O'Keefe and Dostrovsky1971) linked into cognitive maps (O'Keefe & Nadel Reference O'Keefe and Nadel1978; Tolman Reference Tolman1948), grid cells (Moser et al. Reference Moser, Kropff and Moser2008), and head-direction cells (Taube et al. Reference Taube, Muller and Ranck1990). Complementing neural studies are computational models and embodied robots, and it is the fully functional robotic systems (Arleo & Gerstner Reference Arleo and Gerstner2000; Kuipers Reference Kuipers2000; Milford & Wyeth Reference Milford and Wyeth2003; Thrun Reference Thrun, Lakemeyer and Nebel2003) that provide insight for this commentary.

Two approaches can be contrasted for robotic navigational systems: a logically correct approach based on Bayesian reasoning (analogous to deductive reasoning), and one based on a bio-inspired approach that exploits a form of abductive reasoning to constructive a cognitive map. In mobile robots, a key problem is to maintain an estimate of one's current location while exploring and mapping a new environment (called simultaneous localisation and mapping [SLAM]).Given information about localisation (such as a Global Positioning System [GPS]), mapping is a relatively straightforward deductive reasoning problem, and conversely, given a map, localisation is straightforward. However, when both tasks must be solved simultaneously (in the absence of GPS), the errors in each compound. Many locations do not have unique landmarks; apparently unique features of one environment may turn out to be present only transiently or to be shared by other locations. Even recognising a previously visited location at a later time can be challenging. In vision-only SLAM, one of the best-performing systems is the RatSLAM system (Milford Reference Milford2008), inspired by the hippocampal mapping system of the rodent. Initially developed using place cells and head-direction cells, it was discovered early on that the robots also needed something akin to grid cells (although when the model was first developed in 2003, grid cells themselves were yet to be discovered). RatSLAM learns the paths that a robot traverses through its environment and links them into maps. It uses a unique optimisation system that maintains information that is locally consistent, while also estimating a global map.

If a location is considered a “conclusion” in a mapping task, and features of the environment are considered “arguments to support that conclusion,” then systems that are effective at navigation are of necessity abductive reasoners. Maps are constructed by using locations for which there is evidence, and evidence is retained when it is useful for localisation. Maps and their evidence need to be mutually reinforcing to be useful. The hippocampus has been linked to many aspects of cognition as well as spatial memory. Argumentative reasoning may well be the latest of its exapted abilities.

References

Arleo, A. & Gerstner, W. (2000) Spatial cognition and neuro-mimetic navigation: A model of hippocampal place cell activity. Biological Cybernetics 83(3):287–99.CrossRefGoogle Scholar
Kuipers, B. (2000) The spatial semantic hierarchy. Artificial Intelligence 119 (1–2):191233.CrossRefGoogle Scholar
Milford, M. (2008) Robot navigation from nature: Simultaneous localisation, mapping, and path planning based on hippocampal models. Springer-Verlag.Google Scholar
Milford, M. & Wyeth, G. (2003) Hippocampal models for simultaneous localisation and mapping on an autonomous robot. In: Proceedings of the Australasian Conference on Robotics and Automation, Brisbane, Australia. Available at: http://www.araa.asn.au/acra/acra2003/papers/35.pdf.Google Scholar
Moser, E. I., Kropff, E. & Moser, M. B. (2008) Place cells, grid cells, and the brain's spatial representation system. Annual Review of Neuroscience 31:6989.CrossRefGoogle ScholarPubMed
Newell, A. & Simon, H. A. (1956) The logic theory machine: A complex information processing system. IRE Transactions on Information Theory IT-2(3):6179.CrossRefGoogle Scholar
O'Keefe, J. & Dostrovsky, J. (1971) The hippocampus as a spatial map: Preliminary evidence from unit activity in the freely-moving rat. Brain Research 34(1):171–75.CrossRefGoogle ScholarPubMed
O'Keefe, J. & Nadel, L. (1978) The hippocampus as a cognitive map. Oxford University Press.Google Scholar
Peirce, C. S. (1931–35) Collected papers of Charles Sanders Peirce. Harvard University Press.Google Scholar
Taube, J. S., Muller, R. U. & Ranck, J. B. Jr. (1990) Head-direction cells recorded from the postsubiculum in freely moving rats. I. Description and quantitative analysis. Journal of Neuroscience 10(2):420–35.CrossRefGoogle Scholar
Thrun, S. (2003) Robotic mapping: A survey. In: Exploring artificial intelligence in the new millennium, ed. Lakemeyer, G. & Nebel, B., pp. 136. Morgan Kaufmann.Google Scholar
Tolman, E. C. (1948) Cognitive maps in rats and men. Psychological Review 55(4):189208.CrossRefGoogle ScholarPubMed