Codebook

Seed codes carry the dissertation's variables; emergent codes were discovered during coding. Click any code for its node view.

AG — AI agency (79 applications)

Parent: how the source treats AI agency, autonomy, moral status of systems

AG-UNDERSTANDING emergent Whether the AI system genuinely understands values/intentions/causal structure (strong alignment) or merely pattern-matches value-language (weak alignment) - b… 26 / 15 src
AG-AGENTIC Explicitly addresses multi-step, tool-using, delegated-authority agentic AI 22 / 14 src
AG-MORAL-CON Argues AI systems cannot be moral agents (the dissertation's position) 20 / 15 src
AG-CONSTRAINED emergent The position that AI can be subject to / guided by normative principles (morally constrained) without being a moral agent that bears praise/blame (morally acco… 7 / 5 src
AG-COLLECTIVE emergent Effects of many agentic systems interacting: collective autonomy, multi-agent risk, society-level consequences not visible in the dyadic user-agent frame 3 / 2 src
AG-PRE-AGENTIC Argument assumes single-shot/narrow systems; unclear it survives agentic AI 1 / 1 src
AG-MORAL-PRO Argues AI systems can bear moral agency/responsibility (the Luke position) 0 / 0 src

GAP — Gap markers (30 applications)

Parent: gaps this source leaves that the dissertation can occupy

GAP-DESC-ONLY Reports attitudes/practices without deriving or defending normative claims — the is/ought gap Howard flags 14 / 9 src
GAP-AGENTIC-UNTESTED Pre-2023 framework never revisited for agentic systems — the dissertation's opening 8 / 6 src
GAP-NO-EMPIRICS Conceptual argument with no engagement with what stakeholders actually think 8 / 7 src

NF — Normative framework (41 applications)

Parent: which normative-ethical framework the source invokes or presupposes

NF-CONTRACT Source appeals to contractualism, public reason, or overlapping consensus 14 / 7 src
NF-PLURAL-OTHER Pluralist or hybrid framework not reducible to the above 13 / 10 src
NF-KANT Source applies Kantian or rule-deontological machinery 6 / 5 src
NF-CONSEQ Source frames alignment or responsibility in consequentialist terms 5 / 5 src
NF-VIRTUE Source applies virtue ethics or character-based evaluation 2 / 2 src
NF-ROSS Source invokes Ross-style pluralism of prima facie duties 1 / 1 src
NF-NONE Source discusses alignment/responsibility with no articulated normative framework 0 / 0 src

RL — Responsibility locus (101 applications)

Parent: where the source locates responsibility for AI-mediated outcomes

RL-INST Responsibility located in institutional or regulatory structures (Ferretti-style) 34 / 22 src
RL-DEV Responsibility assigned to developers/designers 17 / 13 src
RL-LEGITIMACY emergent Whether those affected have political (democratic-procedural) reason to accept AI outputs: deliberation, meaningful voting, transparency of input-to-policy inf… 16 / 12 src
RL-CONTROL emergent Control/controllability as a condition on alignment or responsibility: the ability to achieve intended/warranted outcomes and intervene, distinct from mere int… 11 / 7 src
RL-DIST Distributed/networked account (Dignum-adjacent) across actors 11 / 9 src
RL-EPISTEMIC emergent Responsibility/liability allocation depends on what an actor understands or can foresee about the system - interpretability and explanation tools therefore RED… 6 / 5 src
RL-SYS Responsibility (or its gap) attributed to the system 3 / 2 src
RL-USER Responsibility assigned to users or deploying organizations 3 / 2 src

TU — Thesis use (85 applications)

Parent: where in the dissertation this material lands

TU-METAETH Feeds the metaethical framework chapter (Ross/Gibbard/convergentism) 31 / 20 src
TU-METHOD Feeds the xphi/corpus methodology chapter 29 / 18 src
TU-LITREV Lit-review positioning only 20 / 16 src
TU-HEALTH Feeds the Health domain chapter 3 / 3 src
TU-IMMIG Feeds the Immigration domain chapter 1 / 1 src
TU-WORK Feeds the Work/Employment domain chapter 1 / 1 src

VC — Value conception (114 applications)

Parent: what the source takes 'values' to be in value alignment

VC-PREF Values operationalized as (revealed/stated/idealized) preferences 33 / 15 src
VC-INTRA-VALUE emergent A single alignment-target value (e.g. autonomy) fragments into conflicting specifications, so alignment requires choosing between rival interpretations WITHIN … 25 / 18 src
VC-PROC Values as outputs of a fair procedure rather than a substantive theory (Gabriel's move) 23 / 9 src
VC-THICK emergent Alignment targets analyzed via the thick/thin concept distinction (Williams/Väyrynen): thin values carry evaluative force without descriptive justification; th… 11 / 10 src
VC-OBJ Values treated as objective goods independent of preferences 9 / 6 src
VC-ROLE emergent Alignment target = the normative ideals/criteria appropriate to the AI system's social role or function (assistant, scribe, interviewer...), rather than anyone… 8 / 6 src
VC-LOCI emergent Source uses or engages Gabriel's taxonomy: instructions / expressed intentions / revealed preferences / informed preferences / interests / values as candidate … 4 / 4 src
VC-EXPRESS Values as expressed attitudes or norms (Gibbard-adjacent) 1 / 1 src

Categories (synthesis layer)

The moral-agency question

Whether AI systems understand, deliberate, or bear moral agency - the four-tradition denial, the constrained-not-accountable settlement, and the empirical instability record

AG-UNDERSTANDING 26 AG-MORAL-CON 20 AG-CONSTRAINED 7 AG-MORAL-PRO 0
The agentic shift

Multi-step, delegated, personalised systems as the stress test every pre-agentic framework fails or defers: collective effects, containment-vs-governance responses, the agency-misuse tradeoff

AG-AGENTIC 22 GAP-AGENTIC-UNTESTED 8 AG-COLLECTIVE 3 AG-PRE-AGENTIC 1
Bridging empirical and normative

The is/ought self-criticism, the missing empirical layer, and the machinery for legitimate bridging: filtration, reflective equilibrium, convergence, measurement

TU-METAETH 31 TU-METHOD 29 GAP-DESC-ONLY 14 GAP-NO-EMPIRICS 8 TU-MWRE 5
Plural, thick, incommensurable values

Value pluralism as datum and design constraint: intra-value fragmentation, thick concepts, objective-list elements, role-indexed standards, and the weighing machinery they demand

VC-INTRA-VALUE 25 NF-PLURAL-OTHER 13 VC-THICK 11 VC-OBJ 9 VC-ROLE 8 NF-KANT 6 NF-CONSEQ 5 NF-ROSS-PF 4 NF-VIRTUE 2 NF-ROSS 1 VC-EXPRESS 1
The rise and fall of preferentism

Preference-based alignment as the field default and the multi-front critique that has broken it: thin-concept, incommensurability, sycophancy, endogeneity, formal Goodhart results

VC-PREF 33 VC-LOCI 4
The procedural turn and its limits

Fair-process/contractualist successors to preferentism (Gabriel program, deliberative and bargaining variants) and their failure points: encoding gap, stakeholder scoping, expert-corrective regress

VC-PROC 23 RL-LEGITIMACY 16 NF-CONTRACT 14
Responsibility: affirmed, not allocated

The responsibility question across loci (developer/user/institution/system/distributed) and conditions (control, epistemic access, legitimacy) - affirmed as shared everywhere, allocated nowhere

RL-INST 34 RL-DEV 17 RL-CONTROL 11 RL-DIST 11 RL-EPISTEMIC 6 RL-SYS 3 RL-USER 3