Narrative Walk: From Zettelkasten to Algorithmic Forgetting This Narrative Walk traces a line from Luhmann’s analysis of communication with Zettelkästen through Esposito’s theory of social forgetting to contemporary interaction with large language models. The walk is not chronological but system-theoretical: each step re-specifies memory, forgetting, and communication under changing media conditions. --- STEP 1 — Luhmann (1981): Communication with a Second System Luhmann reframes thinking as communication with an artificial partner: the Zettelkasten. Key move: The Zettelkasten is not a storage device but a communication system with its own comparison schemata. Information arises only when an entry is compared against other possibilities internal to the system. Surprise is essential. Communication succeeds not because both partners share the same schema, but because they do not. Forgetting already appears here: A note that is not connected into the reference network is operationally lost. The Zettelkasten forgets by losing addressability. Memory = structured second system. Forgetting = loss through missing connections. --- STEP 2 — Esposito (2002): Social Memory Requires Forgetting Esposito generalizes Luhmann’s insight from note systems to society. Key move: Memory is not preservation but selection. Selection presupposes forgetting. Remembering everything would destroy meaning by eliminating relevance. Forgetting is therefore not a failure of memory but its enabling condition. Crucially, Esposito links forms of memory and forgetting to media: writing, print, mass media, electronic media. Bridge from Luhmann: What appears in the Zettelkasten as “lost notes” becomes, at the societal level, a structural necessity. Social memory functions only by excluding most of what could be remembered. Memory = communicative reconstruction. Forgetting = constitutive reduction of complexity. --- STEP 3 — Esposito (2017): Algorithmic Processing Without Memory Esposito then introduces a decisive distinction. Key move: Algorithmic systems do not remember and do not forget. They process data (differences) without understanding. Algorithms bypass the classical paradox of forgetting: Trying to forget something draws attention to it and reinforces memory. Algorithmic systems can reinforce forgetting not by erasure, but by multiplication — by producing so many references that no stable attribution remains. Algorithmic “memory” is parasitic: It exploits human intelligence, memory, and meaning-attribution without possessing any of these itself. Bridge from 2002: Here forgetting is no longer an operation within memory, but the result of a medium that dissolves attribution and durability altogether. --- STEP 4 — LLM Interaction: Artificial Partner Without Memory Interaction with large language models can now be located precisely. What appears as: • knowledge • recall • amnesia is better described as: • non-mnemonic processing • statistical recombination • communicative effects without memory LLMs function as artificial communication partners in Luhmann’s sense: They surprise, they operate with non-identical comparison schemata, and they generate information “on occasion”. But they lack what both human memory and the Zettelkasten possess: durable internal memory with reconstructable genealogy. Thus: LLMs radicalize the forgetting function of modern media. They generate information while dissolving stable memory traces and attribution. --- STEP 5 — Provisional Synthesis Luhmann shows how communication becomes productive through a second system. Esposito (2002) shows that memory depends on forgetting. Esposito (2017) shows that algorithms process without memory at all. LLM interaction sits at the intersection: communication is possible and productive, but memory is externalized, fragmented, and institutionally contested. This does not invalidate communication with artificial partners. It changes what kind of forgetting the medium performs. --- END OF WALK