Hostile Scaffolding Literature (4E Cognitive Science)
Status: PHILOSOPHICAL GROUNDING FOR THE THREAT MODEL.
What this literature provides
A growing critique within 4E cognitive science challenges the "dogma of harmony" (Aagaard 2021) -- the pervasive tendency in extended/embedded cognition to portray human-technology relations as cooperative. Key contributions:
Mind invasion (Slaby 2016). Institutional scaffolding that "hacks" affectivity against an agent's prior orientations. Paradigm cases: toxic corporate environments using digital technology to create "presence bleed," prisons imposing alien norms. Defining feature: the agent is unwilling; the scaffolding contravenes their existing preferences and self-avowed goals.
Hostile scaffolding framework (Timms & Spurrett 2023). Formal definition: external structures deliberately deployed and maintained by one agent at the expense of another. Distinguished from incompetent scaffolding (broken alarm clock) by the presence of both victims and beneficiaries.
Hostile affective technologies (Spurrett 2024). Extends mind invasion to technology that sculpts embodied stance toward the world. Casino environments as paradigm case: lack of windows, free drinks, adaptive gambling machines, fluid mechanisms all conspire to keep gamblers engaged.
Techno-wantonness (White 2025). Argues that "mind invasion" is misapplied to adaptive technology like casinos and social media. Introduces the "techno-wanton": an agent who temporarily displays wanton-like characteristics (Frankfurt 1971) due to engagement with scaffolding that panders to existing first-order preferences while degrading second-order volition. Three hallmarks: (1) low-friction engagement, (2) endless novelty/uncertainty weaponization, (3) internal trigger formation.
Adversarial inference (Bruineberg 2023). Models recommendation algorithms as active inference agents observing user actions and generating content predicted to drive further engagement. When user and algorithm goals misalign, the dynamic becomes adversarial. Links social media to gambling via shared exploitation of anticipatory uncertainty resolution.
Adaptive scaffolding and autonomy (Fige-Talamanca 2024). Identifies recommendation algorithms as "boundary-blurrers" that scaffold vulnerability. White (2025) argues these cases are better described as techno-wantonness than mind invasion.
The mind invasion / techno-wantonness distinction
This distinction maps directly onto AgentSee's threat model:
| Feature | Mind invasion | Techno-wantonness |
|---|---|---|
| Agent's stance | Unwilling | Initially willing |
| Mechanism | Imposes alien preferences | Panders to existing preferences |
| Friction | Increases friction (monitoring, prohibition) | Reduces friction (frictionless engagement) |
| Second-order volition | May remain intact (performative compliance) | Degraded (evaluative governance bypassed) |
| Paradigm case | Prison, toxic workplace | Social media, casino, adaptive algorithms |
| AgentSee relevance | Addressed by I2 (volitional control preserved) | Primary design target: A6, I1, I3, I4 |
Techno-wantonness is the more insidious threat because it exploits the agent's own preferences rather than opposing them. The agent does not experience themselves as a victim until the degradation is well underway.
What this literature does NOT provide
- Neurobiological mechanism. No account of HOW preference-pandering technology degrades evaluative governance at the neural level. AgentSee's answer: catecholamine-mediated PFC suppression (Arnsten 2009, 2015) provides the mechanism. Techno-wanton states correspond to measurable shifts in PFC online/offline status.
- Engineering specification. The literature identifies what goes wrong but not how to build systems that go right. AgentSee provides: control topology (D1), capacity-as-objective (D2), caring constraints (D3), two-layer machine (D4).
- Closed-loop defense. No proposal for real-time state estimation and stabilization to counter techno-wantonness as it occurs. AgentSee's regime table and measurement stack address this.
- Caring governance framework. The hostile scaffolding literature identifies harmful systems but does not specify what a supportive system's normative orientation should be. AgentSee's Mayeroff-grounded caring axiom (A5) and invariants (I1-I6) provide this.
- The exploitative/supportive contrast. White distinguishes mind invasion from techno-wantonness but does not formally contrast exploitative and supportive architectures in control-theoretic terms. See exploitative-technology-design.md.
Positioning statement
The hostile scaffolding literature in 4E cognitive science provides the philosophical articulation of what AgentSee's architecture defends against. White's techno-wantonness concept describes, in philosophical terms, exactly the threat that AgentSee's anti-exploitation axiom (A6) and invariants (I1-I6) are engineered to prevent. The literature's gap -- no neurobiological mechanism, no engineering specification, no closed-loop defense, no caring governance -- is precisely what AgentSee fills. The relationship is complementary: the philosophy identifies the threat; AgentSee provides the defense.