Luciano Floridi (2008), Minds and Machines 18(3), 303–329 (online 15 August 2008; September 2008 issue) — DOI-10.1007/s11023-008-9113-7
Key takeaways
- The paper keeps levels as an epistemological method and drops them as ontology. “Nature does not know about LoAs either” (p. 35).
- The method’s work is done by definitions: observable, LoA, behaviour, moderated LoA and gradient of abstractions (§2). An LoA fixes the range of questions that can be meaningfully asked and answered in principle (p. 18). Adopting an LoA commits a theory to types, and endorsing the model it generates commits it to tokens (pp. 20–21).
- Levels of organisation are ontological and levels of explanation are pragmatic. LoAs underlie both, and the three are not to be confused (§4.1). On this account Marr’s, Pylyshyn’s and Dennett’s tripartite schemes are GoAs that leave their ontological commitment hidden (p. 28).
- Explicit LoAs are offered as giving pluralism without relativism (§4.3) and realism without descriptivism (§4.4). Both are stated more than argued.
Standing in the record
Filed on 2026-10-10 at the owner’s request, with the line of Floridi’s work
on artificial agency that Floridi 2025 (LIT-897) builds on. This is the
method that line uses. Floridi says he and Sanders “were forced to develop
the method of abstraction when we encountered the problem of defining the
nature of agents (natural, human and artificial)” in Floridi & Sanders 2004
(LIT-898) (p. 35). This paper states the method in general form and
refines it, for instance by relating values of types as well as observables
(fn. 8). That is the extends. The earlier and longer joint statement,
Floridi & Sanders 2004, “The Method of Abstraction” (Yearbook of the
Artificial, Peter Lang), is not held here.
It is held here and not in the anthology because it is philosophical method, with no instruction for machine-learning practice. The anthology held none of this batch’s works when checked at commit d8b5ba5.
The record already leans on this method without having held its source. NOTE-099 reads Floridi’s informational structural realism (LIT-152), whose §2.2 restates these definitions and which this paper lists as one of the method’s applications (p. 35, as “Floridi [2004b]”). NOTE-130 reads Floridi’s informational account of personal identity (LIT-136), which assumes the method to deflate diachronic identity. NOTE-172 reads Watson (LIT-187), who builds on “Floridi’s method of levels of abstraction (2008)”. NOTE-700 is the close reading. It placed the work: Active. It is worth reading as the source statement those readings assume. It is a method with definitions, not a result, and its guard against relativism ranks LoAs only once a purpose is fixed.