Roguelite Labs

Personalized Systems

1,218w · 6 min · concept, design, cognition, ai

Personalized Systems

A personalized cognitive system adapts its scaffolding to the user's internal cognitive architecture — amplifying where the user is strong, scaffolding where they are developing, and compensating temporarily where there are current limitations. Unlike customization (the user adapts the tool) or personalization-as-preference (remembering your settings), cognitive personalization means the system understands and adapts to how you think.

The distinction between preference personalization and cognitive personalization is not semantic. Remembering your preferred color scheme or your time zone is a trivial feature. Adapting your system's scaffolding to your working memory capacity, your knowledge gaps, your preferred representational schemas, and your reasoning patterns — that is a different kind of system, with different capabilities and different risks.

The Licklider Vision

J.C.R. Licklider's "Man-Computer Symbiosis" (1960) is the foundational document. He envisioned computers that would learn a person's problem-solving patterns, anticipate their informational needs, and serve as cognitive partners that grew more effective as the relationship developed. This is not a UI that remembers your preferences — it is a system that models your cognitive architecture and adapts its behavior to extend it most effectively.

Licklider's framing was deliberately symbiotic: the system changes as the human changes, and the human changes as the system becomes more capable. The relationship has a developmental arc. Over time, the system builds a model of the human that enables qualitatively different kinds of support; the human builds internal capacities that make qualitatively different demands of the system.

The extended-cognition framework makes this precise: an ideal personalized system adapts to maximize coupling quality between the user's internal cognitive architecture and the system's external processing capacity. Cognitive personalization is the design practice of optimizing that coupling over time.

The PKM Connection — Second Brain

Tiago Forte's Building a Second Brain (2022) and the PKM (Personal Knowledge Management) movement represent extended-cognition made explicit and increasingly mainstream. The second brain: an external system (notes, links, summaries) that mirrors and extends your associative memory, in the spirit of Bush's memex.

Current PKM systems (Obsidian, Roam, Notion) are largely passive stores — they hold information in structures you build. They are personalized only to the extent that the user has personally structured them. The personalization work falls entirely on the user, which limits who benefits: people with the time, skill, and inclination to systematically externalize and link their knowledge compound most from these tools.

The next generation would be active: systems that understand your knowledge structure, surface connections you haven't noticed, and scaffold retrieval in ways that build your internal models over time. The shift from passive second brain to active cognitive partner. Andy Matuschak's mnemonic medium is the clearest current prototype of what this looks like — a reading environment that builds the reader's knowledge into long-term memory through the reading experience, rather than leaving the retention work entirely to the reader.

The FTR Personalization Argument

From future-time-reference: speakers of different FTR language backgrounds bring different default temporal representational schemas to cognitive work. Strong-FTR languages grammatically obligate future-state encoding on every future-referring utterance; weak-FTR languages do not. This means that strong-FTR speakers have had more systematic practice in explicit future-state specification simply from normal language use — their linguistic environment has trained that representational operation more thoroughly.

A personalized system aware of a user's FTR background could scaffold future-state specification differently — externalizing the temporal structure that strong-FTR grammar obligates, for users whose linguistic background hasn't built that internal schema as strongly. This is personalization as adaptive FTR prosthetic: building the cognitive scaffolding that different linguistic environments distribute unequally.

This argument generalizes. Linguistic environment trains cognitive patterns. A user who grew up in an environment that trains high-context communication, or that encodes spatial relations relative to cardinal directions rather than egocentric reference frames, brings different default cognitive schemas to a shared tool. Personalization that accounts for these structural differences is not accommodation of preference — it is compensation for the uneven cognitive substrate that language environments produce.

Personalization vs. Customization

The distinction matters for design and for ethics.

Customization requires the user to know their own cognitive patterns and articulate them. High user control. High effort. Produces personalization only as sophisticated as the user's self-model. Most users have limited and inaccurate self-models of their cognitive patterns. Customization systems built for sophisticated users fail the users who most need cognitive scaffolding.

Personalization requires the system to infer cognitive patterns from behavior and adapt without explicit configuration. Lower user effort. Lower direct user control. The inferred model may be wrong — and systems optimized for engagement rather than cognitive growth will infer the wrong things. Higher privacy stakes: the information required for effective cognitive personalization is also the information that enables cognitive manipulation.

This is not a solvable tradeoff — it is a tension that should be held explicitly in design. The system that most effectively adapts to your cognitive architecture is also the system that most effectively exploits it. The difference between a cognitive prosthetic and a behavioral manipulation engine is the objective the optimization is pointed at.

The Dependency Problem

If a personalized system adapts to your cognitive patterns, you may develop cognitive patterns adapted to the system. This is beneficial if the system is building better internal architecture — if the scaffolding, over time, gets internalized, restructuring your cognitive capacity. It is concerning if the system is substituting for cognitive operations you would otherwise develop.

Vygotsky's framework is precise here: scaffolding is appropriate when it enables operations in the zone of proximal development — operations you can perform with support but not yet independently. The scaffold should dissolve as competence is internalized. A scaffold that is maintained indefinitely either signals that internalization isn't happening or that the system isn't designed to allow it.

A personalized system that makes you more effective only while using it is dependency, not augmentation. The design question — from assistive-technology — is whether the system is a scaffold or a substitute. The personalized system should periodically reduce scaffolding to test whether the user has internalized the scaffolded operations, not maintain scaffolding indefinitely.

Key Researchers and Practitioners

J.C.R. Licklider — "Man-Computer Symbiosis" (1960). The founding vision of computers as adaptive cognitive partners that learn user patterns over time.

Tiago Forte — Building a Second Brain (2022). PKM as explicit extended cognition practice. The most widely adopted implementation of the second brain concept.

Andy Matuschak — active PKM, mnemonic medium, internalization design. The most careful thinker on the difference between storing knowledge and building internal cognitive architecture. His critique: most PKM tools accumulate content without building the user's capacity to think.

Wendy Mackay (INRIA) — appropriation: users modify tools to fit their cognitive practices, going beyond the designer's intent. This is a form of user-driven personalization that reveals what the designed personalization missed.

Vannevar Bush — the memex, associative indexing as personal cognitive architecture. The precursor: a system organized around how a specific mind associates, not around a generic taxonomy.

Related

extended-cognition · tools-for-thought · future-time-reference · assistive-technology · ephemeral-software

Sources

linked from
Ephemeral SoftwareUniversal Design for Cognition