Podcast

This episode is part of a series — click "See More" for all the episodes.


Central Problem

How can perception, understood as a process of data acquisition rather than knowledge in the full sense, generate meaning and ultimately empirical knowledge? The chapter addresses two interconnected difficulties: (1) if knowledge is accounted information, how do we explain perceptual and testimonial knowledge, given that perception and testimony seem to require no explicit “account”? (2) If perception and testimony are reduced to “data providers,” how can raw data become semantically meaningful without presupposing already-given meanings?

The problem is rooted in the tension between an inflated view of perception (which equates it directly with knowledge) and a deflated view (which reduces it to mere sensory input). Floridi seeks a middle path: perception provides “constraining affordances” that cognitive agents creatively repurpose to generate non-natural meanings. This raises the “symbol grounding problem” (SGP): how can a formal system acquire intrinsic semantics without presupposing already-available semantic resources?

Main Thesis

The central thesis is that human beings are “natural-born data hackers”: cognitive agents who repurpose natural data and signals for epistemic, communicative, and semantic purposes that transcend their original natural meanings. This capacity for “semantic hacking” explains how non-natural (conventional) meanings emerge from natural meanings without violating the “Data Processing Theorem” (DPT), according to which data processing cannot increase mutual information.

Perception as data provider: Perception is not knowledge in the full sense, but the process through which agents acquire first-hand data about the environment. To qualify as knowledge, perceptual information must pass the “Phaedrus’ Test” (Platonic test): the agent must be able to answer questions about the why and how of their perceptions.

Testimony as transfer: Testimony transfers already-formed and meaningful information, but does not generate knowledge in itself. To qualify as knowledge, it must pass the “Parrot’s Test” (Cartesian test): receiving true information is not the same as knowing it if one cannot do anything else with it.

Praxical solution to the SGP: Through Action-based Semantics (AbS), agents with a two-module architecture (M1 operational, M2 meta-level) can semantically ground symbols to their sensorimotor interactions with the environment.

Retro-fitness: The correspondence between perception and reality is not passive representation but “retro-fitness”: perceptual models of the world are outputs of data hacking, not inputs, and compete evolutionarily for correctness and fitness for purpose.

Historical Context

The chapter fits within the post-cognitivist debate on the nature of perception and knowledge, developing themes from the neo-Kantian philosophy of Helmholtz (1878), who already interpreted sensations as “signs” rather than “images” of reality. Floridi takes up the Gricean distinction (1957) between natural and non-natural meaning (meaningNN), criticizing naturalist attempts (Skyrms 2010) to completely reduce the latter to the former.

The scientific context includes Shannon’s information theory, the data processing theorem (DPT), and contemporary neuroscience showing the brain as a proactive “predictor” rather than a passive “mirror” (Nobre et al. 2007). The approach opposes both naive realism and relativism, proposing a relational “third way” inspired by Platonic metaxy.

On the artificial intelligence front, the failure of the classical AI project is interpreted as confirmation of the insufficiency of syntactic processing to generate genuine semantics (“semantics-from-syntax fallacy”).

Philosophical Lineage

flowchart TD
    Plato --> Floridi
    Kant --> Helmholtz
    Helmholtz --> Floridi
    Leibniz --> Floridi
    Grice --> Floridi
    Shannon --> Floridi
    Wittgenstein --> Floridi
    Peirce --> Floridi
    Popper --> Floridi
    Cassirer --> Floridi

    class Plato,Kant,Helmholtz,Leibniz,Grice,Shannon,Wittgenstein,Peirce,Popper,Cassirer,Floridi internal-link;

Key Thinkers

ThinkerDatesMovementMain WorkCore Concept
Helmholtz1821-1894Neo-KantianismHandbook of Physiological OpticsPerceptions as signs, not images
Grice1913-1988Analytic PhilosophyMeaning (1957)Natural vs non-natural meaning
Shannon1916-2001Information TheoryA Mathematical Theory of CommunicationMutual information, DPT
Cassirer1874-1945Neo-KantianismSubstance and FunctionMathematical structuralism
Popper1902-1994Critical RationalismObjective KnowledgeWorld 3, subjectless knowledge

Key Concepts

ConceptDefinitionRelated to
Data hackingRepurposing natural data/signals for non-natural semantic purposesFloridi, Constructionism
Symbol Grounding ProblemHow formal symbols acquire intrinsic meaning without presupposing semanticsHarnad, AI
Action-based SemanticsTheory of meaning based on internal states correlated with actionsFloridi, Pragmatics
Phaedrus’ TestCriterion: knowing requires being able to answer questions about one’s knowledgePlato, Epistemology
Parrot’s TestCriterion: receiving true information is not the same as knowing itDescartes, Testimony
Data Processing TheoremData processing does not increase mutual information: I(X;Y) ≥ I(X;Z)Shannon, Information Theory
Constraining affordancesData as enabling constraints for semantic constructionFloridi, Perception
Retro-fitnessPerceptual models are outputs of data hacking, competing for correctnessFloridi, Constructionism
DedomenaData “in the wild” before cognitive interpretationFloridi, Ontology
Natural-born data hackersDefinition of humans as creative repurposers of meaningsFloridi, Cognitive Science

Authors Comparison

ThemeFloridiGriceHelmholtz
Nature of perceptionData provider, constraining affordancesPresupposes semanticsSigns, not images
MeaningNatural → non-natural via hackingNatural vs non-natural (meaningsNN)Sign relation, not pictorial
World-mind relationshipConstructionism, retro-fitnessNot systematically thematizedNeo-Kantianism, mediation
KnowledgeAccounted informationIntentional communicationInferred regularities

Influences & Connections

Summary Formulas

  • Floridi: Perception provides data that agents creatively “hack” to generate non-natural meanings; knowledge is accounted information, not mere reception of data.
  • Helmholtz: Sensations are signs of external reality, not images; they represent regularities, not resemblances.
  • Grice: Non-natural meaning (meaningsNN) cannot be reduced to natural meaning through mere causal processing or conditioning.
  • Data Processing Theorem: Data processing does not generate new information; I(X;Y) ≥ I(X;Z) implies that non-natural semantics requires something beyond processing.

Timeline

YearEvent
1878Helmholtz publishes the theory of sensations as signs
1957Grice distinguishes natural and non-natural meaning
1948Shannon formalizes information theory
1990Harnad formulates the Symbol Grounding Problem
2011Floridi proposes the praxical solution to SGP in The Philosophy of Information
2019Floridi develops the “data hacking” hypothesis in The Logic of Information

Notable Quotes

“Humans are natural-born data hackers.” — Floridi

“Our sensations are effects brought forth in our organs by means of exterior causes […] a sign need have no similarity of any sort whatever with that of which it is the sign.” — Helmholtz

“We are not evolution’s finest moment, the peak of the process, some kind of Über-animal, but Nature’s beautiful glitch.” — Floridi