Research Environment II · Unreal Engine (C++ / Blueprints)
Untamed Shadows
Open Social System · Inter-group Dynamics · Cross-engine Validation
Research Instrument Context
Untamed Shadows is the cross-validation environment for the PhD simulation framework. Its purpose is not to replicate Prison Sandbox in a different setting — it is to test whether the agent framework generalises to a structurally different social topology: open territory instead of closed space, inter-group dynamics instead of individual relationships, faction-level trust propagation instead of personal trust. The Unreal implementation validates engine-agnosticism.
Why This Environment — Cross-Validation Design Rationale
Open Territory — Social Topology Contrast
Prison Sandbox uses a closed, bounded population. Untamed Shadows uses open wilderness territory with permeable faction boundaries. This structural contrast is the point: if the agent framework produces coherent emergent narrative in both, it demonstrates generalisability across fundamentally different social geometries. A framework that only works in one topology is not a framework — it is a genre-specific hack.
Faction-level Trust — Group Propagation
Prison Sandbox models trust as an individual-level construct. Untamed Shadows models trust at the faction level — information credibility is mediated by group membership and inter-faction history. This allows the framework to test whether RQ2's trust modeling mechanisms scale from personal relationships to institutional ones, which is a key generalisability question.
Act Structure — Consequence at Scale
Choices made in Act 2 permanently restructure Act 3 — faction alliances, available territories, and agent relationships. This implements consequence permanence (Pillar III) at a narrative structural level rather than a moment-to-moment tactical level, allowing the human evaluation phase to distinguish between micro-scale and macro-scale irreversibility effects on perceived narrative coherence.
Engine Change — Framework Validation
The agent simulation layer runs identically beneath Unreal Engine's C++/Blueprints architecture as beneath Unity's C#. This is not a platform preference — it is a methodological validation. The cross-engine design is itself a contribution (Section 6 of the proposal): it demonstrates the framework is a generalizable research tool, not a game-specific implementation.
Environments Contrast — What Each Environment Tests
| Dimension | Prison Sandbox (Env. I) | Untamed Shadows (Env. II) |
|---|---|---|
| Social topology | Closed bounded population, fixed agent count | Open territory, permeable faction boundaries |
| Trust unit | Individual-to-individual personal trust | Faction-level group trust propagation |
| Information flow | Rumor chains through close proximity | Territorial intelligence across faction domains |
| Consequence scope | Moment-to-moment: NPC death, route closure | Structural: Act 2 choices reshape Act 3 world |
| Engine | Unity · C# · top-down 2D/3D | Unreal Engine 4 · C++ / Blueprints · third-person |
| Primary validation | RQ1: information reliability → network structure | RQ4: framework generalises across social topologies |
Faction Architecture — Social Agent Groups
The Wolves
Trust: Territory-bound
Share information freely within pack, withhold from all others. High intra-group reliability, near-zero inter-group. Models tight social trust with sharp external boundary.
The Pumas
Trust: Transactional
Information exchanged for territorial access. Reliability is contingent on reciprocity. Models market-mediated trust — honest when it benefits them, misleading otherwise.
The Lions
Trust: Hierarchical
Information flows downward accurately; upward information is filtered and aggrandised. Models how hierarchy distorts information reliability through social propagation.
Human Research
Trust: Observer
Holds accurate environmental data but its own agenda — scientific access — shapes what it discloses. Models institutional information control as a reliability variable.
Research Variable Mapping
| Inter-faction information gaps | Different factions hold incompatible information about shared territory. The player cannot verify without traversal. Models information reliability variance at the group level. | Pillar I · RQ1 |
| Faction trust accumulation | Trust built through side missions and sustained presence. Trust with one faction affects credibility with rivals. Models how group-level trust propagates and conflicts. | Pillar II · RQ2 |
| Act 2 → Act 3 consequence | Structural narrative bifurcation based on accumulated faction choices. Tests whether macro-scale irreversibility affects perceived narrative coherence differently than micro-scale. | Pillar III · RQ3 |
| Cross-engine framework layer | Agent simulation API runs identically under Unreal and Unity. This architectural constraint is a direct test of RQ4 — whether a parameterized framework predicts narrative coherence across implementations. | RQ4 |
Implementation Status
Untamed Shadows is designed as a research instrument; Unreal implementation is Phase 1 of the PhD. The design documentation above demonstrates that both environments have been methodologically specified — structural decisions, variable mappings, and cross-validation rationale are established before the PhD begins.