Johannes Jaeger, Anna Riedl, Alex Djedovic, John Vervaeke & Denis Walsh
"Naturalizing relevance realization: why agency and cognition are fundamentally not computational"
(resubmitted with minor revisions to Frontiers in Psychology)
(preprint published in Dec 2023)
______
The way organismic agents come to know the world, and the way algorithms solve problems, are fundamentally different. The most sensible course of action for an organism does not simply follow from logical rules of inference. Before it can even use such rules, the organism must tackle the problem of relevance. It must turn ill-defined problems into well-defined ones, turn semantics into syntax. This ability to realize relevance is present in all organisms, from bacteria to humans. It lies at the root of organismic agency, cognition, and consciousness, arising from the particular autopoietic, anticipatory, and adaptive organization of living beings. In this paper, we show that the process of relevance realization is beyond formalization. It cannot be captured completely by algorithmic approaches. This implies that organismic agency (and hence cognition as well as consciousness) are at heart not computational in nature. Instead, we show how the process of relevance is realized by an adaptive and emergent triadic dialectic (a trialectic), which manifests as a metabolic and ecological-evolutionary co-constructive dynamic. This results in a meliorative process that enables an agent to continuously keep a grip on its arena, its reality. To be alive means to make sense of one's world. This kind of embodied ecological rationality is a fundamental aspect of life, and a key characteristic that sets it apart from non-living matter.
"Naturalizing relevance realization: why agency and cognition are fundamentally not computational"
(resubmitted with minor revisions to Frontiers in Psychology)
(preprint published in Dec 2023)
______
The way organismic agents come to know the world, and the way algorithms solve problems, are fundamentally different. The most sensible course of action for an organism does not simply follow from logical rules of inference. Before it can even use such rules, the organism must tackle the problem of relevance. It must turn ill-defined problems into well-defined ones, turn semantics into syntax. This ability to realize relevance is present in all organisms, from bacteria to humans. It lies at the root of organismic agency, cognition, and consciousness, arising from the particular autopoietic, anticipatory, and adaptive organization of living beings. In this paper, we show that the process of relevance realization is beyond formalization. It cannot be captured completely by algorithmic approaches. This implies that organismic agency (and hence cognition as well as consciousness) are at heart not computational in nature. Instead, we show how the process of relevance is realized by an adaptive and emergent triadic dialectic (a trialectic), which manifests as a metabolic and ecological-evolutionary co-constructive dynamic. This results in a meliorative process that enables an agent to continuously keep a grip on its arena, its reality. To be alive means to make sense of one's world. This kind of embodied ecological rationality is a fundamental aspect of life, and a key characteristic that sets it apart from non-living matter.
Johannes Jaeger
Assembly theory: what it does and what it does not do
Journal of Molecular Evolution 92: 87-92 (March 2024)
(based on a blog post published in Oct 2023)
______
A recent publication in Nature has generated much heated discussion about evolution, its tendency towards increasing diversity and complexity, and its potential status above and beyond the known laws of fundamental physics. The argument at the heart of this controversy concerns assembly theory, a method to detect and quantify the influence of higher-level emergent causal constraints in computational worlds made of basic objects and their combinations. In this short essay, I briefly review the theory, its basic principles and potential applications. I then go on to critically examine its authors’ assertions, concluding that assembly theory has merit but is not nearly as novel or revolutionary as claimed. It certainly does not provide any new explanation of biological evolution or natural selection, or a new grounding of biology in physics. In this regard, the presentation of the paper is starkly distorted by hype, which may explain some of the outrage it created.
Assembly theory: what it does and what it does not do
Journal of Molecular Evolution 92: 87-92 (March 2024)
(based on a blog post published in Oct 2023)
______
A recent publication in Nature has generated much heated discussion about evolution, its tendency towards increasing diversity and complexity, and its potential status above and beyond the known laws of fundamental physics. The argument at the heart of this controversy concerns assembly theory, a method to detect and quantify the influence of higher-level emergent causal constraints in computational worlds made of basic objects and their combinations. In this short essay, I briefly review the theory, its basic principles and potential applications. I then go on to critically examine its authors’ assertions, concluding that assembly theory has merit but is not nearly as novel or revolutionary as claimed. It certainly does not provide any new explanation of biological evolution or natural selection, or a new grounding of biology in physics. In this regard, the presentation of the paper is starkly distorted by hype, which may explain some of the outrage it created.
Johannes Jaeger, James DiFrisco & Andrea Loettgers
"Beyond Networks: Explaining Dynamics in the Natural and Social Sciences"
(forthcoming book chapter)
(preprint published in Dec 2023)
______
This paper presents a broad criticism of network modeling in both the natural and the social sciences, and a study of where its shortcomings originate historically, following and earlier paper by James and myself. Here, we extend our earlier argument from the specific discipline of evolutionary-developmental biology (or evo-devo) to a broad range of network models in neuroscience, cell & developmental biology, ecology, and economics. We trace the origin of network models not only to graph theory but also lattice models (eg the Ising model) in condensed-matter physics in the 1920s, showing that their application to other disciplines follows a complex entangled pattern of model template transfers.
Model templates capture an intertwinement of mathematical structure, computational tools, and theoretical concepts that depict a general mechanism applicable to any field displaying a particular pattern of interactions. Network models are successful model templates not because they provide accurate representations of the target systems to which they are applied, but because of their generic ontology & associated conceptual framework that can be used to model a wide range of different systems.
We provide extensive examples, starting with connectionist models in (cognitive) neuroscience, arguing that an excessive focus on network structure obscures important phenomena linked to the dynamics of the system. We then criticize network approaches in cell & developmental biology, focusing on scale-free networks, network motifs, & Davidson's genomic regulatory networks. Again, an excessive focus on structure severely limits our outlook because structure does not determine function. Reverse-engineered networks based on a connectionist formalism may provide a compromise between abstraction of network structure & a focus on dynamical properties of regulatory systems. Further examples come from ecology, where time-invariant models fail to capture dynamic changes in feeding habits during an organism's life cycle, or the non-discrete dynamics of ecological succession and extinction events. Finally, we study the fascinating entanglement of Ising's original lattice model of ferromagnetism and Schelling's economic model of segregation in urban environments, showing that template transfer occurs in non-obvious ways, with unexpected feedback between disciplines.
We frame the entire complex of problems that is illustrated by these examples as a mismatch in the possibility spaces generated by static network models & differential-equation models based on dynamical systems theory. Simply put, the possibility space of static network models is too restricted to capture many relevant phenomena in the domains of application of network models. The restrictions mainly stem from the intrinsic nature of the underlying model templates. In turn, the possibility space of dynamical models also fails to capture many relevant phenomena that depend on the fact that biological systems (and those based on them) constantly alter their organization based on the intrinsic dynamics of the system.
To conclude: as attractive & versatile as network models are, their power is also their downfall: by abstracting away the dynamics of a system, we are bound to restrict ourselves to small subsets of possible phenomena to be studied & their plausible scientific explanations.
"Beyond Networks: Explaining Dynamics in the Natural and Social Sciences"
(forthcoming book chapter)
(preprint published in Dec 2023)
______
This paper presents a broad criticism of network modeling in both the natural and the social sciences, and a study of where its shortcomings originate historically, following and earlier paper by James and myself. Here, we extend our earlier argument from the specific discipline of evolutionary-developmental biology (or evo-devo) to a broad range of network models in neuroscience, cell & developmental biology, ecology, and economics. We trace the origin of network models not only to graph theory but also lattice models (eg the Ising model) in condensed-matter physics in the 1920s, showing that their application to other disciplines follows a complex entangled pattern of model template transfers.
Model templates capture an intertwinement of mathematical structure, computational tools, and theoretical concepts that depict a general mechanism applicable to any field displaying a particular pattern of interactions. Network models are successful model templates not because they provide accurate representations of the target systems to which they are applied, but because of their generic ontology & associated conceptual framework that can be used to model a wide range of different systems.
We provide extensive examples, starting with connectionist models in (cognitive) neuroscience, arguing that an excessive focus on network structure obscures important phenomena linked to the dynamics of the system. We then criticize network approaches in cell & developmental biology, focusing on scale-free networks, network motifs, & Davidson's genomic regulatory networks. Again, an excessive focus on structure severely limits our outlook because structure does not determine function. Reverse-engineered networks based on a connectionist formalism may provide a compromise between abstraction of network structure & a focus on dynamical properties of regulatory systems. Further examples come from ecology, where time-invariant models fail to capture dynamic changes in feeding habits during an organism's life cycle, or the non-discrete dynamics of ecological succession and extinction events. Finally, we study the fascinating entanglement of Ising's original lattice model of ferromagnetism and Schelling's economic model of segregation in urban environments, showing that template transfer occurs in non-obvious ways, with unexpected feedback between disciplines.
We frame the entire complex of problems that is illustrated by these examples as a mismatch in the possibility spaces generated by static network models & differential-equation models based on dynamical systems theory. Simply put, the possibility space of static network models is too restricted to capture many relevant phenomena in the domains of application of network models. The restrictions mainly stem from the intrinsic nature of the underlying model templates. In turn, the possibility space of dynamical models also fails to capture many relevant phenomena that depend on the fact that biological systems (and those based on them) constantly alter their organization based on the intrinsic dynamics of the system.
To conclude: as attractive & versatile as network models are, their power is also their downfall: by abstracting away the dynamics of a system, we are bound to restrict ourselves to small subsets of possible phenomena to be studied & their plausible scientific explanations.
Johannes Jaeger
"Artificial Intelligence is Algorithmic Mimicry: Why Artificial "Agents" Are Not (and Won't Be) Proper Agents"
NBDT 1-21 (February 2024)
(preprint published in Jun 2023)
______
What is the prospect of developing artificial general intelligence (AGI)?
I investigate this question by systematically comparing living and algorithmic systems, with a special focus on the notion of "agency."
There are three fundamental differences to consider:
(1) Living systems are autopoietic, that is, self-manufacturing, and therefore able to set their own intrinsic goals, while algorithms exist in a computational environment with target functions that are both provided by an external agent.
(2) Living systems are embodied in the sense that there is no separation between their symbolic and physical aspects, while algorithms run on computational architectures that maximally isolate software from hardware.
(3) Living systems experience a large world, in which most problems are ill-defined (and not all definable), while algorithms exist in a small world, in which all problems are well-defined.
These three differences imply that living and algorithmic systems have very different capabilities and limitations. In particular, it is extremely unlikely that true AGI (beyond mere mimicry) can be developed in the current algorithmic framework of AI research.
Consequently, discussions about the proper development and deployment of algorithmic tools should be shaped around the dangers and opportunities of current narrow AI, not the extremely unlikely prospect of the emergence of true agency in artificial systems.
"Artificial Intelligence is Algorithmic Mimicry: Why Artificial "Agents" Are Not (and Won't Be) Proper Agents"
NBDT 1-21 (February 2024)
(preprint published in Jun 2023)
______
What is the prospect of developing artificial general intelligence (AGI)?
I investigate this question by systematically comparing living and algorithmic systems, with a special focus on the notion of "agency."
There are three fundamental differences to consider:
(1) Living systems are autopoietic, that is, self-manufacturing, and therefore able to set their own intrinsic goals, while algorithms exist in a computational environment with target functions that are both provided by an external agent.
(2) Living systems are embodied in the sense that there is no separation between their symbolic and physical aspects, while algorithms run on computational architectures that maximally isolate software from hardware.
(3) Living systems experience a large world, in which most problems are ill-defined (and not all definable), while algorithms exist in a small world, in which all problems are well-defined.
These three differences imply that living and algorithmic systems have very different capabilities and limitations. In particular, it is extremely unlikely that true AGI (beyond mere mimicry) can be developed in the current algorithmic framework of AI research.
Consequently, discussions about the proper development and deployment of algorithmic tools should be shaped around the dangers and opportunities of current narrow AI, not the extremely unlikely prospect of the emergence of true agency in artificial systems.
Johannes Jaeger
"Ontogenesis, Organization, and Organismal Agency"
(forthcoming book chapter)
(preprint published in Jun 2022)
______
At first sight, the empirical study of ontogenesis and the theoretical study of organismal agency seem to have little in common. This essay discusses why this initial impression is incorrect.
First of all, ontogenesis and agency are indirectly connected at the level of the whole organism, since they are co-dependent on the peculiar organization that characterizes living systems. While ontogenesis is constrained by its own requirement to maintain living organization in the form of organizational closure throughout the life cycle, agency is grounded in the same phenomenon of organizational continuity.
Second, cellular agency contributes more directly to important processes of multicellular development in organisms with multiple levels of organization. This leads to a view of ontogenesis that emphasizes agency and variation in the underlying cellular dynamics, focusing on stability and reproducibility of ontogenetic processes as its main explanatory targets.
We examine how these insights can help us bridge the explanatory gap between reductionist mechanistic empirical approaches and theoretical considerations of the organization of the whole organism. We conclude that both approaches are best used in a complementary manner. Only by contextualizing ontogenetic mechanisms in the larger context of the evolving life cycle, will we gain a true understanding of their functionality and evolution.
"Ontogenesis, Organization, and Organismal Agency"
(forthcoming book chapter)
(preprint published in Jun 2022)
______
At first sight, the empirical study of ontogenesis and the theoretical study of organismal agency seem to have little in common. This essay discusses why this initial impression is incorrect.
First of all, ontogenesis and agency are indirectly connected at the level of the whole organism, since they are co-dependent on the peculiar organization that characterizes living systems. While ontogenesis is constrained by its own requirement to maintain living organization in the form of organizational closure throughout the life cycle, agency is grounded in the same phenomenon of organizational continuity.
Second, cellular agency contributes more directly to important processes of multicellular development in organisms with multiple levels of organization. This leads to a view of ontogenesis that emphasizes agency and variation in the underlying cellular dynamics, focusing on stability and reproducibility of ontogenetic processes as its main explanatory targets.
We examine how these insights can help us bridge the explanatory gap between reductionist mechanistic empirical approaches and theoretical considerations of the organization of the whole organism. We conclude that both approaches are best used in a complementary manner. Only by contextualizing ontogenetic mechanisms in the larger context of the evolving life cycle, will we gain a true understanding of their functionality and evolution.
Andrea Roli, Johannes Jaeger, and Stuart A. Kauffman
"How Organisms Come to Know the World: Fundamental Limits on Artificial General Intelligence"
Frontiers in Ecology & Evolution 9: 806283 (January 2022).
(preprint published in October 2021)
______
Artificial intelligence has made tremendous advances since its inception about seventy years ago. Self-driving cars, programs beating experts at complex games, and smart robots capable of assisting people that need care are just some among the successful examples of machine intelligence. This kind of progress might entice us to envision a society populated by autonomous robots capable of performing the same tasks humans do in the near future. This prospect seems limited only by the power and complexity of current computational devices, which is improving fast.
However, there are several significant obstacles on this path. General intelligence involves situational reasoning, taking perspectives, choosing goals, and an ability to deal with ambiguous information. We observe that all of these characteristics are connected to the ability of identifying and exploiting new affordances — opportunities (or impediments) on the path of an agent to achieve its goals. A general example of an affordance is the use of an object in the hands of an agent.
We show that it is impossible to predefine a list of such uses. Therefore, they cannot be treated algorithmically. This means that “AI agents” and organisms differ in their ability to leverage new affordances. Only organisms can do this.
This implies that true AGI is not achievable in the current algorithmic frame of AI research.
It also has important consequences for the theory of evolution. We argue that organismic agency is strictly required for truly open-ended evolution through radical emergence.
We discuss the diverse ramifications of this argument, not only in AI research and evolution, but also for the philosophy of science.
"How Organisms Come to Know the World: Fundamental Limits on Artificial General Intelligence"
Frontiers in Ecology & Evolution 9: 806283 (January 2022).
(preprint published in October 2021)
______
Artificial intelligence has made tremendous advances since its inception about seventy years ago. Self-driving cars, programs beating experts at complex games, and smart robots capable of assisting people that need care are just some among the successful examples of machine intelligence. This kind of progress might entice us to envision a society populated by autonomous robots capable of performing the same tasks humans do in the near future. This prospect seems limited only by the power and complexity of current computational devices, which is improving fast.
However, there are several significant obstacles on this path. General intelligence involves situational reasoning, taking perspectives, choosing goals, and an ability to deal with ambiguous information. We observe that all of these characteristics are connected to the ability of identifying and exploiting new affordances — opportunities (or impediments) on the path of an agent to achieve its goals. A general example of an affordance is the use of an object in the hands of an agent.
We show that it is impossible to predefine a list of such uses. Therefore, they cannot be treated algorithmically. This means that “AI agents” and organisms differ in their ability to leverage new affordances. Only organisms can do this.
This implies that true AGI is not achievable in the current algorithmic frame of AI research.
It also has important consequences for the theory of evolution. We argue that organismic agency is strictly required for truly open-ended evolution through radical emergence.
We discuss the diverse ramifications of this argument, not only in AI research and evolution, but also for the philosophy of science.
Johannes Jaeger
"The Fourth Perspective: Evolution and Organismal Agency"
In: "Organization in Biology" ed. Matteo Mossio, Springer, Berlin (2024).
(preprint published in February 2021)
______
This essay examines the deep connections between biological organization, agency, and evolution by natural selection.
Its central argument is that the basic unit of evolution is not a genetic replicator, but a complex hierarchical life cycle, or reproducer. It shows that the self-manufacturing capabilities of reproducers are a necessary precondition for evolvability, and proposes an extended and disambiguated set of minimal conditions for evolution by natural selection — including new or revised principles of heredity, variation, and ontogenesis (broadly defined as the acquisition of the capacity to reproduce).
The requirement for continued maintenance of self-manufacturing organization throughout the life cycle and across generation suggests that all evolvable systems are agents (or contain agents among their components). This means that we ought to take agency seriously, if we aim to obtain an organism-level theory of evolution in the original spirit of Darwin's struggle for existence.
Such understanding must rely on an agential perspective on evolution, complementing and succeeding existing structural, functional, and processual approaches. This essay sketches a tentative outline of such an agential perspective, and presents a survey of methodological and conceptual challenges that will have to be overcome if we are to properly implement it.
"The Fourth Perspective: Evolution and Organismal Agency"
In: "Organization in Biology" ed. Matteo Mossio, Springer, Berlin (2024).
(preprint published in February 2021)
______
This essay examines the deep connections between biological organization, agency, and evolution by natural selection.
Its central argument is that the basic unit of evolution is not a genetic replicator, but a complex hierarchical life cycle, or reproducer. It shows that the self-manufacturing capabilities of reproducers are a necessary precondition for evolvability, and proposes an extended and disambiguated set of minimal conditions for evolution by natural selection — including new or revised principles of heredity, variation, and ontogenesis (broadly defined as the acquisition of the capacity to reproduce).
The requirement for continued maintenance of self-manufacturing organization throughout the life cycle and across generation suggests that all evolvable systems are agents (or contain agents among their components). This means that we ought to take agency seriously, if we aim to obtain an organism-level theory of evolution in the original spirit of Darwin's struggle for existence.
Such understanding must rely on an agential perspective on evolution, complementing and succeeding existing structural, functional, and processual approaches. This essay sketches a tentative outline of such an agential perspective, and presents a survey of methodological and conceptual challenges that will have to be overcome if we are to properly implement it.
