orientation · ~2 min read
How to read this
If you just landed and you're looking at a chart full of dots and chips and amber lines, this page is the short walkthrough. No methodology, no jargon — just what you're looking at and how to use it.
Each placed by a panel — a human prior plus nine language models — with every score the median of the panel and its disagreement kept visible rather than averaged away.
1. The compass
The home page is a two-dimensional plot. Every dot is one political actor — a party or a politician — placed by where they sit on two axes. Hover any dot to see who they are; click to read their full profile.
You can filter by country, by political family, by lens, and by what kind of actor you want to see. The point of the compass is comparison: how far apart are Likud and Labor? Where does Hezbollah sit relative to Amal? Does the Muslim Brotherhood overlap with anyone in Tunisia?
2. The two axes
Horizontal: economic. Left means more state allocation, public ownership, redistribution. Right means more market mechanisms, private ownership, lower taxes. A welfare-state social democrat sits on the left; a privatize-everything liberal sits on the right.
Vertical: social. Top means more authoritarian — strong executive, restricted civil liberties, traditional moral order enforced. Bottom means more libertarian — civil liberties expanded, individual autonomy prioritized, the state stepping out of personal life.
Both axes run from −10 to +10. A score of 0 means roughly neutral or mixed. A score of ±10 means a strong, consistent position. Most actors sit somewhere in the middle of one axis and have a stronger position on the other.
Beyond the two main axes, the Issues page lists every other axis we score on — economic governance, military doctrine, civil liberties, foreign alignment, the Palestine question, and others. Those are single-issue rankings; the compass is the headline summary.
3. Reading a dot
Hover any dot to see the name, country flag, and exact coordinates. Click to open the actor's page. On the actor page you'll find:
- Their position on every axis (not just the two main ones) as a row of small bars
- A short description of the party or person
- Their public statements and platform documents, where we have them
- Citations for every claim we make about their position
Outlines around dots mean different things. A solid ring marks a party currently in government. A dashed, hollow dot is historical — dissolved, merged away, or no longer competing. A red outline means the party is outlawed or banned. A plain dot with no ring is simply a party that isn't in government.
4. Confidence and verification
Politics is messy. We try to be honest about how sure we are.
Every party and politician page shows a small chip indicating when they were last reviewed. "First-pass reviewed" means the position has been written once and checked once. We're upfront that a second-pass external review hasn't happened yet — it's a planned step before we publish anything as a research paper.
The Pulse feed (and the timeline view) carries a similar idea for events: confirmed events are solid, reported events are amber, rumored events are dashed, speculative is dotted. The labels are deliberately conservative. If you see "rumored," treat it as rumor.
5. Lenses
A party can say one thing in its manifesto and vote another way in parliament. Lenses are a way to look at that difference rather than averaging it out.
- Composite — the default. Our best single-number summary across all evidence.
- Declared — what they say in manifestos, official platforms, public statements.
- Behavioral — how they actually vote, what they actually do in government.
Switch lenses with the toggle at the top of the compass. Not every actor has all three — many only have the composite, which is what the toggle falls back to. The cases worth seeking out are the ones where declared and behavioral diverge; that gap is often the real story.
6. Where the panel disagrees — and why that's the point
Nine models rarely score an actor identically, and Tayyar never averages the disagreement away. The gaps are often the most interesting thing on the page.
On regime stance, the panel splits across the entire −10 to +10 range for Hayat Tahrir al-Sham: most models place it at the revolutionary floor, one near the top. It had just gone from sanctioned insurgency to governing Syria, and models trained at different moments are scoring different facts. The split isn't an error — it's the panel flagging an actor caught mid-transition.
This is the design, not a defect. A cell where the raters agree is marked settled; one where they diverge is marked contested and shown as a range, not a point. The ratings page lays out every rater's score for every cell, and findings zooms out to the patterns.
7. Where things live
- Compass
- The two-axis plot of all actors, filterable. Its own page now.
- Map
- A geographic view. Countries shaded by their mean position.
- Issues
- Every axis we score on, with rankings per axis.
- Pulse
- Recent regional events, sorted newest first.
- Timeline
- The same events, year by year, oldest first. Better for "what happened when."
- Bills
- Specific pieces of legislation we track, with positions of parties on each.
- Compare
- Put two parties (or politicians, or countries) side by side.
- Findings
- What the data shows once you zoom out — correlations, clusters, outliers.
- Methodology
- How we score. The recipe.
- Limitations
- What we're not doing, and what could go wrong. Read this if you want to cite us.
- Dataset
- Download the whole thing as CSV or JSON. CC BY-NC-SA 4.0.
Where to next
That covers how to read the data. If a score or a term is still unclear, that's likely on us — tell us and we'll rewrite the explanation.