Algorithmic structure of experience and the unfolding argument

Ruffini, Giulio; Lopez-Sola, Edmundo; Vohryzek, Jakub

doi:10.31234/osf.io/7nbsw

Cited by 4 publications

(6 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In Ruffini et al [2022b], we point out one important weakness of the argument, based on the existence of the Chomsky-Schützenberger computational hierarchy [Chomsky & Schützenberger, 1963;Fitch, 2014]-where FNNs sit essentially at the bottom. For example, we note that sufficiently large RNNs can be made Turing complete (v. Siegelmann & Sontag [1995] and the discussion in Ruffini [2016]), while it is fairly intuitive that the set of all FNNs is not Turing complete.…”

Section: Philosophy Of Sciencementioning

confidence: 99%

AIT Foundations of Structured Experience

Ruffini¹,

Lopez-Sola²

2022

J. AI. Consci.

Self Cite

View full text Add to dashboard Cite

In this paper, we present a unifying framework to study consciousness based on algorithmic information theory (AIT). We take as a premise that “there is experience” and focus on the requirements for structured experience ([Formula: see text]) — the spatial, temporal, and conceptual organization of our first-person experience of the world and of ourselves as agents in it. Our starting point is the insight that access to good models — succinct and accurate generative programs of world data — is crucial for homeostasis and survival. We hypothesize that the successful comparison of such models with data provides the structure to experience. Building on the concept of Kolmogorov complexity, we can associate the qualitative aspects of [Formula: see text] with the algorithmic features of the model, including its length, which reflects the structure discovered in the data. Moreover, a modeling system tracking structured data will display dimensionality reduction and criticality features that can be used empirically to quantify the structure of the program run by the agent. KT provides a consistent framework to define the concepts of life and agent and allows for the comparison between artificial agents and [Formula: see text]-reporting humans to provide an educated guess about agent experience. A first challenge is to show that a human agent has [Formula: see text] to the extent they run encompassing and compressive models tracking world data. For this, we propose to study the relation between the structure of neurophenomenological, physiological, and behavioral data. The second is to endow artificial agents with the means to discover good models and study their internal states and behavior. We relate the algorithmic framework to other theories of consciousness and discuss some of its epistemological, philosophical, and ethical aspects.

show abstract

Section: Philosophy Of Sciencementioning

confidence: 99%

AIT Foundations of Structured Experience

Ruffini¹,

Lopez-Sola²

2022

J. AI. Consci.

Self Cite

View full text Add to dashboard Cite

show abstract

“…These may be realized in artificial or biological systems in a holistic, distributed, hierarchical manner involving brain networks and complex multistable dynamics. As postulated in [44], the theoretical framework exists to formalize the notion of such modules. For example, for an agent to fulfill its regulative objectives, it necessitates a modeling engine of the parameters it seeks to control.…”

Section: Algorithmic Agent In Mddmentioning

confidence: 99%

The algorithmic agent perspective and computational neuropsychiatry: from etiology to advanced therapy in major depressive disorder

Ruffini,

Castaldo,

Lopez-Sola

et al. 2024

Preprint

View full text Add to dashboard Cite

Major Depressive Disorder is a complex and heterogeneous condition impacting millions of individuals globally. Computational neuropsychiatry is poised to deliver significant breakthroughs through mechanistic modeling of this condition. To address the multifactorial character of Major Depressive Disorder, we frame our analysis within the Algorithmic or Kolmogorov Information Theory of consciousness, where agents interact with the world driven by an objective function that evaluates {\em valence}. In this context, depression is defined as a state characterized by persistently low valence. This view suggests that depression may originate from inaccurate world models (leading to negative bias), a dysfunctional objective function (anhedonia), deficient planning (cognitive deficits), or unfavorable environmental conditions --- potentially linking the model with proposed depression biotypes. Building upon existing literature, we bridge algorithmic, dynamical systems and neurobiological concepts, emphasizing the role of plasticity --- the ability of the agent to update elements such as their world model or objective function --- in maintaining psychological health. This perspective connects and expands on emerging concepts in the REBUS and CANAL perspectives, such as {\em landscape flattening and canalization}. Finally, we explore how the synergies between brain stimulation, psychotherapy and plasticity-enhancing compounds such as psychedelics can be used to repair neural circuits, and how these therapies can be optimized using personalized computational models - {\em neurotwins}.

show abstract

“…We do not need to rely on that theory again to interpret the predicted effect. Instead, we carry the experiments on ourselves (or on other volunteers) to determine if we feel a difference in experience [for the volunteers, we can rely on their introspective reports, provided that the above-mentioned appropriate circumstances occur; see also ( Ruffini et al, 2022 )]. Obviously, such experiments, ideally employing a multitude of methods, can provide converging evidence that either increases or decreases our degree of belief in the hypothesis under study (note that we qualify this to human consciousness), and we do not need to assume a theory to arrive at this result.…”

Section: Examining the Philosophical Premises Of The Ua (P1 P3 P4)mentioning

confidence: 99%

“…While the UA proponents claim that input-output (i.e., behavioral) data are the primary evidence for consciousness science and that purely phenomenological data, which are not translated into some public marker, are not scientific data at all, we maintain that behavioral data dreams) depend on this sort of self-sustained activation states that are possible only for RNs, and thus appear (in the absence of input) outside the reach of FFNs. Second, there are independent reasons to reject P2: research within the field of neural computing (Siegelmann and Sontag, 1995;Cabessa and Siegelmann, 2012) indicates that FFN and RN differ drastically in terms of their computational power: while the former are far below Turing-computation, the latter can exceed it (see (Ruffini et al, 2022) for a recent and detailed discussion; see also Usher (2021) for an illustration showing that apparently equivalent RN and FFN, are not equivalent when tested with perturbations, and replies by Herzog et al, 2022). like introspective reports can only be valuable heuristics to be used in experimental settings, but they cannot be taken at face value.…”

Section: The Scientific Significance Of the First-person Perspectivementioning

confidence: 99%

“…This is particularly important for our aims, because many conscious states (e.g., dreams) depend on this sort of self-sustained activation states that are possible only for RNs, and thus appear (in the absence of input) outside the reach of FFNs. Second, there are independent reasons to reject P2: research within the field of neural computing ( Siegelmann and Sontag, 1995 ; Cabessa and Siegelmann, 2012 ) indicates that FFN and RN differ drastically in terms of their computational power: while the former are far below Turing-computation, the latter can exceed it (see ( Ruffini et al, 2022 ) for a recent and detailed discussion; see also Usher (2021) for an illustration showing that apparently equivalent RN and FFN, are not equivalent when tested with perturbations, and replies by Herzog et al, 2022 ).…”

mentioning

confidence: 99%

See 1 more Smart Citation

When philosophical nuance matters: safeguarding consciousness research from restrictive assumptions

Usher,

Negro,

Jacobson

et al. 2023

Front. Psychol.

View full text Add to dashboard Cite

In this paper, we revisit the debate surrounding the Unfolding Argument (UA) against causal structure theories of consciousness (as well as the hard-criteria research program it prescribes), using it as a platform for discussing theoretical and methodological issues in consciousness research. Causal structure theories assert that consciousness depends on a particular causal structure of the brain. Our claim is that some of the assumptions fueling the UA are not warranted, and therefore we should reject the methodology for consciousness science that the UA prescribes. First, we briefly survey the most popular philosophical positions in consciousness science, namely physicalism and functionalism. We discuss the relations between these positions and the behaviorist methodology that the UA assumptions express, despite the contrary claim of its proponents. Second, we argue that the same reasoning that the UA applies against causal structure theories can be applied to functionalist approaches, thus proving too much and deeming as unscientific a whole range of (non-causal structure) theories. Since this is overly restrictive and fits poorly with common practice in cognitive neuroscience, we suggest that the reasoning of the UA must be flawed. Third, we assess its philosophical assumptions, which express a restrictive methodology, and conclude that there are reasons to reject them. Finally, we propose a more inclusive methodology for consciousness science, that includes neural, behavioral, and phenomenological evidence (provided by the first-person perspective) without which consciousness science could not even start. Then, we extend this discussion to the scope of consciousness science, and conclude that theories of consciousness should be tested and evaluated on humans, and not on systems considerably different from us. Rather than restricting the methodology of consciousness science, we should, at this point, restrict the range of systems upon which it is supposed to be built.

show abstract

Algorithmic structure of experience and the unfolding argument

Cited by 4 publications

References 15 publications

AIT Foundations of Structured Experience

AIT Foundations of Structured Experience

The algorithmic agent perspective and computational neuropsychiatry: from etiology to advanced therapy in major depressive disorder

When philosophical nuance matters: safeguarding consciousness research from restrictive assumptions

Contact Info

Product

Resources

About