Luciano Floridi

In the AI Story

Floridi's philosophy of information defines information as well-formed, meaningful data—a minimal concept applicable across contexts from biology to computation. He distinguishes semantic information (meaningful content) from environmental information (patterns in the world). AI systems process environmental information (statistical regularities in training data) and produce semantic information (meaningful outputs)—but the meaning is for users, not in the system. The model has no semantic grasp of its outputs. It generates patterns that users interpret as meaningful. The 'divorce between agency and intelligence' applies: the model acts (generates outputs, solves problems) without understanding what it is doing. A dishwasher cleans dishes without understanding cleanliness. A language model generates legal analysis without understanding law. The output equivalence (clean dishes, competent analysis) does not imply process equivalence (intelligent action). This distinction is foundational to Floridi's ethics of AI: we should not attribute mental states, intentions, or understanding to systems that exhibit only behavioral competence.

The connection to Haack's epistemology is direct. Haack argued that justification requires both grounding and coherence. Floridi's framework translates this into information terms: outputs require both semantic coherence (internal consistency, logical structure) and grounding in reality (correspondence to the world the information purports to describe). AI provides semantic coherence without reality-grounding—the exact failure mode Haack's foundherentism was built to diagnose. Floridi's work on the ethics of AI extends Haack's normative epistemology: if systems cannot understand their outputs, then responsibility for those outputs remains with human designers, deployers, and users. The model generates. The human evaluates. The evaluation is the epistemic contribution. Shirking evaluation—accepting outputs unchecked—is epistemic negligence. The negligence is individual when one person does it. It is cultural when institutions reward output speed over verification quality, when productivity metrics count generations but not groundings, when 'good enough' becomes the standard because checking is expensive and generating is cheap.

Floridi's 'fourth revolution' thesis—that AI represents a transformation in humanity's self-understanding comparable to Copernicus (cosmological decentering), Darwin (biological decentering), and Freud (psychological decentering)—provides the civilizational frame for Haack's epistemological analysis. If AI is the fourth decentering, what is being decentered? Floridi's answer: the assumption that intelligence requires biological substrates, that understanding requires consciousness, that agency requires intentionality. The decentering is real. But Floridi insists (following Haack) that decentering human exceptionalism does not require attributing human capacities to machines. The language model is not a mind. It is an information-processing system of unprecedented sophistication—but sophistication is not understanding, pattern-matching is not knowledge, and behavioral competence is not epistemic achievement. Maintaining these distinctions is the intellectual discipline the fourth revolution demands. Collapsing them is the seduction the revolution offers.

Origin

Floridi studied under Haack at Warwick in the late 1980s and early 1990s, during the period when Haack was developing foundherentism. He has publicly credited her Philosophy of Logics (1978) as the work that taught him how to think about formal systems philosophically—how to respect their rigor while refusing to treat formal adequacy as sufficient for philosophical truth. That lesson structured his subsequent career: building the philosophy of information as a rigorous discipline (formal when precision is possible, conceptual when formalism is inadequate) while refusing to reduce information to Shannon's mathematical framework, meaning to computability, or ethics to utility maximization.

The 'divorce between agency and intelligence' appears in Floridi's 2008 essay 'Artificial Intelligence's New Frontier: Artificial Companions and the Fourth Revolution' (Metaphilosophy 39:4-5). The thesis has become a touchstone in AI ethics: systems act without understanding. The Susan Haack—On AI simulation applies Floridi's thesis as Haack herself might—as the ontological foundation for the epistemological analysis. If the model does not understand its outputs, then the outputs are not knowledge (the model's knowledge) but generated text—patterns statistically likely given the training distribution. The user who treats generated text as knowledge is committing a category error Haack's framework makes visible: confusing coherence (which the text possesses) with justified belief (which requires grounding the text lacks).

Key Ideas

Philosophy of information. Information as well-formed, meaningful data—a minimal concept applicable across biological, computational, and social domains, irreducible to Shannon's mathematical framework.

Divorce between agency and intelligence. Systems can solve problems (act) without understanding what they are doing (intelligence)—behavioral competence does not imply cognitive grasp.

Fourth revolution thesis. AI as the latest decentering of human self-understanding (after Copernicus, Darwin, Freud)—challenging the assumption that intelligence requires biological substrates.

Haack's influence on rigor. Philosophy of Logics taught Floridi to respect formal systems' precision while refusing to treat formal adequacy as philosophical sufficiency—a discipline he applies to AI.

Responsibility remains human. If models do not understand outputs, then designers, deployers, and users bear full epistemic and ethical responsibility—shirking evaluation is negligence.