Symbolic AI (often referred to as Good Old-Fashioned AI, or GOFAI) is the classical, rule-based tradition of artificial intelligence. Knowledge is represented as logic formulas, frames, rules, or knowledge graphs; inference proceeds by theorem proving, search, consistency checking, or the application of production rules. It was the dominant form of AI until the 1990s and is today, under the sign of explainability requirements, again on the rise as a component of neuro-symbolic AI.
Example systems range from the classical expert systems of the 1980s (MYCIN, DENDRAL) through automated planners (STRIPS, PDDL) to modern OWL reasoners such as HermiT, Pellet, or Konclude and the proof assistants Coq, Isabelle, and Lean.
In terms of the ontology of personhood, symbolic AI is a form of artificial intelligence and shares its determination as a tool. The symbols are signs — they refer because a person has assigned them meaning; their chaining in inference rules is program, not an act of understanding.
Ontological classification: Superordinate concept: Artificial Intelligence; siblings: Neural AI, Neuro-Symbolic AI (pairwise disjoint — a concrete system is assigned to exactly one architectural class).
Strengths and limits
Strengths of the symbolic tradition: complete traceability (every inference step is an applied rule), verifiability (consistency checking against the logic), integrability of expert knowledge (domain knowledge is explicitly encoded), data efficiency (no massive training required).
Limits: brittleness in the face of unclear inputs (perception, natural language, noisy sensor data), the knowledge-acquisition problem (rules must be written by humans), scaling problems in highly branching domains, lack of robustness toward inputs outside the modeled world.
Neuro-symbolic AI understands itself as a systematic answer to these limits: the symbolic layer retains explainability, while the neural layer opens up the capacities for perception and language.
Personal-ontological classification
Within the substance-ontological conception of intelligence, symbolic AI is no special case: it appears clearly as a tool and does not claim to be more. It therefore has exclusively derived intentionality: the meaning of every symbol comes from the person who introduced it. The reasoner computes what follows logically — it does not understand what it computes.
This clear tool-nature makes symbolic AI especially unsuspicious for personal-ontological analysis: no one would misread it as a person, a bearer of consciousness, or a simulation of persons. Unlike with large language models, no quasi-personalities arise here as a side effect of the architecture.
Sources: Generated by querying the Personhood ontology. Research as of 7 June 2026 (dossier Neuro-Symbolic AI — Worldwide Research).
Further sources:
- Russell, Stuart & Norvig, Peter (2021): Artificial Intelligence. A Modern Approach. 4th ed. Pearson.
- Brachman, Ronald J. & Levesque, Hector J. (2004): Knowledge Representation and Reasoning. Morgan Kaufmann.
- Glimm, Birte; Horrocks, Ian; Motik, Boris; Stoilos, Giorgos & Wang, Zhe (2014): HermiT: An OWL 2 Reasoner. Journal of Automated Reasoning 53(3): 245—269.