Skip to main content

Towards Autonomous Artificial Agents? Proposal for a Naturalistic Activity-Based Model of (Artificial) Life

  • Chapter
  • First Online:
Blended Cognition

Part of the book series: Springer Series in Cognitive and Neural Systems ((SSCNS,volume 12))

  • 469 Accesses

Abstract

While the contemporary achievements of AI and robotics are indisputable, the issue of autonomy for artificial agents still looms ahead despite technological progress and rich conceptual debates. Drawing on recent theoretical propositions from the enactive approach on autonomy, we first highlight several limitations of what we call an identity-based model. Through the study of four real-life cases, we then not only argue that autonomy cannot be conflated with behavioral self-maintenance or organizational closure, but that it can sometimes violate these conditions. Finally, we propose a naturalistic activity-based model of autonomous agents that emphasizes the importance of norm-establishing processes distributed across an intricacy of milieus.

The animal is immediately one with its life activity. It is not distinct from that activity; it is that activity.

Karl Marx

[...] think only of the ambition of men, and you will wonder at the senselessness of their ways, unless you consider how they are stirred by the love of an immortality of fame. They are ready to run all risks greater far than they would have run for their children, and to spend money and undergo any sort of toil, and even to die, for the sake of leaving behind them a name which shall be eternal.

Diotima of Mantinea, Plato’s Symposium

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Chapter
USD 29.95
Price excludes VAT (USA)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
USD 129.00
Price excludes VAT (USA)
  • Available as EPUB and PDF
  • Read on any device
  • Instant download
  • Own it forever
Hardcover Book
USD 169.99
Price excludes VAT (USA)
  • Durable hardcover edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Notes

  1. 1.

    Our distinction between autonomy and automation is inspired by Ruyer’s cybernetics (Ruyer 1954) and its opposition between mechanism (which is normatively encompassed by an activity) and activity (which is normatively encompassing mechanisms). For further discussion on cybernetics and automation, see Guchet (2010).

  2. 2.

    See Jonas (1968) for a philosophical attempt to ground existentialism in biology.

  3. 3.

    Also called “constructive” and “interactive” in Di Paolo (2009).

  4. 4.

    As Parisi (2014) explains, working on the autonomy of artificial agents outside the framework of a particular given task implies a form of “robotics as science” distinct from (but not necessarily opposed to) “robotics as practical applications”. Interestingly, autonomy (of artificial agents) as a conceptual issue inevitably pertains to the structural issue of scientists’ autonomy. This is something that we think needs to be discussed further by social scientists and engineers within the field of artificial life.

  5. 5.

    A kind of specification issue that has also been discussed on different terms regarding models of “extended-cognition” (Rowlands 2010; Adams and Aizawa 2010)—which cannot in good conscience be conflated with enactivism (Rowlands 2009).

  6. 6.

    These dynamics are often referred to as “environmental dynamics”, but we wish to avoid any confusion with our reframing of the environment’s role in autonomous system.

  7. 7.

    Even though we should mention that some authors have on the contrary insisted on the gap and major differences between what they see as sensorimotor enactivism and autopoietic enactivism (Degenaar and O’Regan 2015).

  8. 8.

    Although as Froese and Ziemke point out in their critique of Parisi’s drive architecture (Parisi 2014), flexibility is not sufficient to instantiate meaning: “[...] consider Parisi’s example of a robot which is provided with two inputs that are supposed to encode its motivational state in terms of hunger and thirst. While it is clear that these inputs play a functional role in generating the overall behavior of the robot, any description of this behavior as resulting from, for example, the robot’s desire to drink in order to avoid being thirsty must be deemed as purely metaphorical at best and misleading at worst. [...] The shift of focus toward sensorimotor loops was an important step in the right direction since it resulted in more robust and flexible systems, but it nevertheless did not fully solve the problem of meaning in AI” (Froese and Ziemke 2009, 472).

  9. 9.

    The endogeneity of the preference could nonetheless be reinforced by taking inspiration of other developmental works, such as the incongruity function in Oudeyer and Kaplan (2007), where a stable environment “gets boring” over time - something that should also generate a transition of preference.

  10. 10.

    “[...] we can re­define habit as “a self­-sustaining pattern of sensorimotor coordination that is formed when the stability of a particular mode of sensorimotor engagement is dynamically coupled with the stability of the mechanisms generating it. [...] a habit can take on a life of its own: it is both the cause and the consequence of its own enactment. This form of recursion makes it possible to understand a mild sense of identity for the habit, a locus of survival and self­-generating persistence” (Barandiaran 2008, 281).

  11. 11.

    “[...] as self-sustaining structures, [habits] are never bad for themselves” (Di Paolo 2009, 18). To which we could ask: what about disgust and boredom for one’s own behavior?

  12. 12.

    “[...] a specific (sensorimotor) normative dimension can be operationalized or measured for this [mentally coherent] kind of network of habits: the viable limits (of disruption, decoupling, etc.) out of which the organization is irreversibly lost. A norm emerges, taking the form of a Kantian imperative or regulatory principle: behave so as to sustain your capacity to behave.” (Barandiaran 2016, 25).

  13. 13.

    For an extensive critique of Jonas’ philosophy of life, see Barbaras (2008, 2010).

  14. 14.

    “What an organism does (both as complex metabolic system and as a natural agency in its world) is to actively seek its own continuation. Those aspects of its interactions that contribute to this natural purpose are seen as intrinsically good according to this self-generated norm. And those aspects that challenge this end, are intrinsically bad” (Di Paolo 2003, 8).

  15. 15.

    This does not rule out the possibility of being affected -- even triggered -- by environmental events. On the contrary: research in ‘autonomous’ (i.e. situated) robotics, and in most computational neuroethology (CNE), focuses specifically on a creature’s reactive responses to environmental cues. Even research that is based on the theory of autopoiesis, which stresses the system’s ability to form (and maintain) itself as a functioning unity, posits that a cell, or an organism, is closely coupled with its environmental surroundings – so much so that they can be regarded as a single system (Maturana and Varela 1980).

  16. 16.

    “[...] “living” media show aspects of operational closure at collective levels without necessarily showing individuality. These include patterns of parental scaffolding, social reproduction structures like Bourdieu’s habitus (1990), epistemic form of niche-construction, stigmergy [...] “down to” the processes that occur in the extra-cellular matrix of multicellular organisms, or in biofilms in bacterial collectives. Such cases put in evidence that the richness of environments as active media has so far been underplayed in the current enactive story. EM has sought to thematize this richness and the enactivist should listen.” (Di Paolo 2009, 20).

  17. 17.

    Jonas explains that “It is a matter of to-be-or-not-to-be whether what is to be done is done” (Jonas 1968, 243) and Di Paolo that “[...] we need a prior identity to which relations originally refer” (Di Paolo 2009, 19).

  18. 18.

    A counter-argument could be made by positing that the octopus was only behaving in such a way because she was driven to by genetic or other deterministic internal conditions. This does not weaken our argumentation for two reasons. First, it is still impossible to make the organizational closure the unified source of normativity because the agent’s activity would still be valued phylogenetically. This issue refers to the tension between autopoietic and Darwinian approaches (see Ruiz-Mirazo et al. 2004). Second, it is highly problematic to fall back on such genetic determinism because on one hand, it is quite rare to find a bijective model between genetic markers and complex behaviors (even with the mediation of environmental constraints) (Oyama et al. 2001), and on the other hand the invocation of adaptive optimality and strategy for reproduction is often plagued by misconceptualization (Amundson 2001; Williams 1966; Abrams 2001) and lack of empirical data (Jamieson 1986; Gould 1984; Williams 1966). It is thus reasonable to describe the sacrifice of the octopus-mother as a normatively specific activity and not only as the mere output of a genetic “programming”.

  19. 19.

    “Specialist tracker dogs have noses which are at least 10 000–100 000 times more sensitive than humans to some odorants, and they seem par­ticularly sensitive to butyric acid” (Sillar et al. 2016, 43). However, what we want to stress here is not the outstanding dog’s sensibility, nor what differentiates its Umwelt from those of other animals, but the simple fact that the effort to be made to avoid leaving any traces in the environment is much bigger than what is required to leaving one. We have to reverse the picture: leaving traces (intentionally or unintentionally) is the rule, and completely avoiding it is impossible.

  20. 20.

    As Turner puts it: “Speaking of interaction between an organism and its environment clearly implies two things. One, of course, is the effect of the environment upon the organism, and the second is the effect of the organism upon the environment. Most of modern environmental physiology is focused on the first (...). Rarely are effects acting the other way considered.” (Turner 2000, 10–11). But rather than endorsing his contentious notion of “extended organisms”, we prefer to say that both organisms and the environment are constituted by the extension of activities.

  21. 21.

    Just like Barbaras explained that plants’ being is one of limited animality (Barbaras 2010, 111), we argue that activities should be understood in reference to the intersecting nature of traps.

  22. 22.

    Research in prey-predator relations have found far-from-equilibrium dynamics of coevolution called “Red Queen Dynamics”, according to which preys and predators see their phenotypes changing indefinitely. This “stochastic” and “polymorphic” model is based on a “limit-cycle” attractor that is the normative center of a never-ending co-constitutive evolution (Dieckmann et al. 1995). Prey and predator, through their activity (harvesting/escaping), are effectively creating the conditions that norm their mutual evolutionary dependency.

  23. 23.

    “[...] this communication [between the agent’s habits and the milieu] is far from being only regulative and adaptive—as if the organism had to follow a law of self-conservation; it gives rise to actions and encounters, it regulates an organized conduct (defense of the territory, search for a mate, education of the progeny) that has a goal, a series of phases and finally a consummatory end” [our emphasis and translation] ( Simondon 2010, 65).

References

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Corentin Chanet .

Editor information

Editors and Affiliations

Rights and permissions

Reprints and permissions

Copyright information

© 2019 Springer Nature Switzerland AG (outside the USA)

About this chapter

Check for updates. Verify currency and authenticity via CrossMark

Cite this chapter

Chanet, C., Eubelen, D. (2019). Towards Autonomous Artificial Agents? Proposal for a Naturalistic Activity-Based Model of (Artificial) Life. In: Vallverdú, J., Müller, V. (eds) Blended Cognition. Springer Series in Cognitive and Neural Systems, vol 12. Springer, Cham. https://doi.org/10.1007/978-3-030-03104-6_10

Download citation

Publish with us

Policies and ethics