Mylfed 24 11 15 Freya Von Doom And Claire Roos New -

The final utility for a candidate branch b is:

[ U(b) = \alpha \cdot \overlineA_b + \beta \cdot E_b + \gamma \cdot C_b ]

The TL then visualises the contribution of each term via a radial glyph displayed in the peripheral HUD. mylfed 24 11 15 freya von doom and claire roos new

DAM augments the MylFed pipeline with two new subsystems:

| Subsystem | Function | Implementation Highlights | |-----------|----------|----------------------------| | Fine‑Grained Agency Tracker (FGAT) | Assigns an agency weight (−1 → +1) to every dialogue act, action, and environmental cue. | Utilises a Bidirectional LSTM trained on a corpus of 12 k annotated interactive scripts (Roos & von Doom, 2023). | | Adaptive Affective Mapper (AAM) | Dynamically calibrates emotion classification per user using Online Transfer Learning. | Starts from a generic CNN (Liu & Picard, 2014) and fine‑tunes on the first 3 minutes of gameplay. | | Transparency Layer (TL) | Generates in‑game “agency hints” (visual glyphs) explaining why a narrative branch was selected. | Based on the Explainable AI (XAI) toolkit Shapley‑Narratives (Kelley et al., 2022). | The final utility for a candidate branch b

Figure 1 (below) depicts the data flow.

[Player Input] → FGAT → NGE → BS → TL → [Narrative Output]
          ↘                ↗
          AAM ← Biosignals (OpenBCI)

| Area | Key Contributions | Relevance to MylFed | |------|-------------------|----------------------| | Procedural Narrative Generation | Façade (Mateas & Stern, 2005); Story‑Flow (Riedl & Young, 2010) | Early inspiration for branching logic | | Affective Computing in Games | Emotion Engine (Bailenson et al., 2012); Affect‑Responsive Narrative (Liu & Picard, 2014) | Basis for biosignal integration | | Player Agency Metrics | Agency Scale (Murray, 1997); Perceived Control Index (Nacke & Lindley, 2010) | Foundations for our agency evaluation | | Ethical Frameworks for Adaptive Systems | Transparency by Design (Kelley & Breazeal, 2019); AI‑Ethics for Games (Baker & O’Brien, 2021) | Guides our discussion section | The TL then visualises the contribution of each

While these works address components of the problem space, none combine real‑time affective feedback, fine‑grained agency quantification, and transparent adaptive narration in a single, open‑source pipeline. DAM therefore fills a critical gap.

| Demographic | N | Age (M ± SD) | Gender | |-------------|---|---------------|--------| | Experienced gamers (≥ 20 h/wk) | 48 | 24.3 ± 3.1 | 28 M / 20 F | | Casual players (≤ 5 h/wk) | 36 | 27.8 ± 4.5 | 18 M / 18 F |

All participants gave informed consent and were compensated £25.