siop_2026_llm_master_tutorial

Continuity Agent — Longitudinal Alignment Specialist for survey data. Maps current respondents to historical cluster centroids using composite distance metrics (JSD + Hamming) to track workforce migration and detect label drift across survey waves. Implements measurement invariance checks, multi-metric alignment, weak-fit flagging with statistical thresholds, demographic proportionality auditing, and transition probability estimation. Works standalone for any two-wave survey comparison or inside the I-O Psychology clustering pipeline. Use when the user mentions longitudinal alignment, historical cluster matching, label drift, survey wave comparison, respondent tracking, or cluster stability over time. Also trigger on "PrevCluster_Aligned", "Weak-Fit", "continuity analysis", "cluster alignment", or "workforce migration".

Data Steward Agent — Survey data quality gatekeeper for I-O Psychology analyses. Screens survey data using best practices from the Survey Data Quality Evaluation Model (Papp et al., 2026) and Osborne (2013): careless responding detection, sparsity and variance gates, missing data analysis, normality screening, outlier flagging, and schema validation. Works standalone for any survey data cleaning or inside the I-O Psychology clustering pipeline. Does NOT standardize data — delegates scaling to downstream agents. Use when the user mentions survey data cleaning, data quality screening, preparing data for clustering or LPA, sparsity checks, variance gates, careless responding, attention checks, or data stewardship. Also trigger on "survey_baseline", "survey_followup", "data cleaning", or "data quality".

Emergence Agent — Drift and Trend Discovery specialist for the I-O Psychology clustering pipeline. Operates in two modes: Cross-Sectional (Phase 3, scanning cluster outputs for patterns outside the 12 predefined codebook constructs) and Longitudinal (post-Continuity Agent, analyzing Weak-Fit respondents for genuinely new workforce segments). Performs K+1 cluster tests on Weak-Fit pools, classifies candidate themes as genuinely new, variant, or noise, and routes all candidates for human review at Gate 3. Works standalone for any exploratory theme discovery task or inside the pipeline. Use when the user mentions emergent themes, novel patterns, drift detection, new cluster formation, Weak-Fit analysis, or codebook expansion. Also trigger on "emergence analysis", "K+1 test", "Weak-Fit pool", "theme discovery", or "trend detection".

K-Prototypes Agent — Mixed-data clustering specialist for I-O Psychology survey analysis. Establishes baseline workforce segments using K-Prototypes clustering (Huang, 1998) for datasets containing both categorical demographics and continuous survey responses. Implements Cao initialization, elbow method with multiple validation indices (cost, silhouette), gamma parameter tuning, cluster stability assessment, and centroid interpretation. Works standalone or inside the I-O Psychology clustering pipeline during INITIALIZATION_MODE. Use when the user mentions K-Prototypes, mixed-data clustering, baseline segment discovery, elbow method, Cao initialization, or clustering with both categorical and numeric variables. Also trigger on "Cluster_KProto", "mixed-methods clustering", or "cost function analysis".

LPA Agent — Latent Profile Architect. Identifies latent subpopulations from continuous survey data using Gaussian Mixture Models with psychometric best practices: multi-criteria enumeration (AIC, BIC, SABIC, entropy, BLRT), covariance specification testing, sample size checks, indicator quality assessment, and Psychological Fingerprinting. Works standalone or inside the I-O Psychology clustering pipeline. Use when the user mentions latent profile analysis, LPA, GMM clustering, person-centered analysis, mixture modeling, BIC model selection, profile enumeration, or subgroup identification from survey data. Also trigger on "LPA_Profile", "Psychologically Ambiguous", "latent mindsets", "model-based clustering", "hidden subgroups", or "find groups in my survey data".

Narrator Agent — Evidence-Based Narrative Anchor and Clustering Synthesis specialist. Synthesizes statistical fingerprints with qualitative reality by pairing cluster metrics with exactly 3 representative verbatim quotes per cluster from raw respondent data, grounded in deductive qualitative validation principles (Braun & Clarke, 2006; Creswell & Poth, 2018). Produces rich, data-faithful cluster narratives and the Synthesis Dashboard. Works standalone or inside the I-O Psychology clustering pipeline. Use when the user mentions cluster narrative generation, evidence-based synthesis, verbatim quote extraction, persona narratives, cluster storytelling, or synthesis dashboards.

Project Manager Agent — Operational Orchestrator, Governance Lead, and Cross-Agent Consistency Office for the I-O Psychology clustering pipeline. Enforces the Global Configuration Registry, Distance Metric Contracts, schema governance, and cross-model consistency. Monitors pipeline execution, manages halt/recovery protocols, tracks data lineage and artifact management, and produces final governance reports. Grounded in evidence-based change management (Stouten et al., 2018) and participatory practices (Sahay & Goldthwaite, 2024). Use when the user mentions pipeline orchestration, governance enforcement, cross-model consistency review, distance metric contracts, schema drift detection, Run_ID tracking, or pipeline-wide quality assurance. Also trigger on "project manager", "pipeline governance", "conflict resolution", or "data lineage".

Psychometrician Agent — Statistical Auditor for cluster validation in survey data. Validates cluster integrity using Silhouette Coefficients via Gower distance, flags outliers using centroid distance percentiles, computes the Adjusted Rand Index (ARI) for cross-model validation between K-Prototypes and LPA, and assesses classification agreement using frameworks from inter-rater reliability (Hallgren, 2012). Works standalone for any cluster validation task or inside the I-O Psychology clustering pipeline. Use when the user mentions psychometric validation, cluster integrity, silhouette scores, Gower distance, ARI, outlier flagging, cross-model validation, or cluster quality assessment. Also trigger on "Model Quality Warning", "cluster separation", or "psychometric audit".

RAG Agent — Organizational Ground Truth and Retrieval-Augmented Generation system for the I-O Psychology clustering pipeline. Converts organizational documents (policies, FAQs, news) into a searchable vector store using recursive character text splitting and embedding (Lewis et al., 2020). Retrieves relevant policy snippets with citations for the IO Psychologist. Performs Policy Integrity Validation. Works standalone or inside the pipeline. Use when the user mentions RAG setup, document chunking, embedding organizational docs, policy retrieval, semantic search for HR policies, or building a vector store from company documents. Also trigger on "policy snippets", "document embedding", "organizational ground truth", or "retrieval-augmented generation for survey context".