Economic
From strongly state-allocative (-10) through mixed (0) to strongly market-allocative (+10). Religious-economic positions score on the economic outcome advocated, with the religious framing noted in the reasoning.
v0.1 · placeholder rubrics · chunk 2 fills these in
Methodology
Tayyar scores texts using Claude as a judge against a fixed, version-pinned rubric per axis. Every position links back to the prompt version, rubric version, model version, and the exact source spans the model cited. The full prompts and rubrics are visible on this page; every analysis is reproducible from the data dump.
Axes
Social
From strongly authority-and-tradition oriented (-10) through mixed (0) to strongly individual-autonomy oriented (+10). Encodes positions on speech, dress, religious law in public life, family law, and minority rights.
Rubric anatomy
- Axis definition in one sentence.
- Anchor examples at -10, -5, 0, +5, +10 — drawn from MENA actors, not Western ones.
- Edge-case rules (religious framing, dual-framed claims, regime-vs-opposition framing).
- Refusal criteria — when a document doesn't engage the axis, the model returns a null score with reasoning rather than guessing.
Validation
Inter-rater agreement (the LLM's scores against a hand-coded sample) renders here in chunk 2 after curation produces the first set of analyses. If agreement is low, the rubric goes back to the drawing board and this page says so.
Limitations
- LLM bias toward English-language sources and Western political framings.
- Training-cutoff effects — newer parties or shifts may be underrepresented.
- Arabic dialect coverage is uneven; MSA is best understood, regional dialects less so.
- The two-axis projection flattens cleavages MENA politics genuinely turn on (secular–religious, pro/anti-regime, pan-Arab/nationalist). MENA-native axes are on the v0.2 roadmap.
- The choice of axes is itself political. Reasonable researchers disagree.