Scoring Methodology
The Scoreboard tracks whether AI developments are moving power toward concentration, dispersion, or something more mixed.
It is not a prediction machine.
It is not a claim of mathematical certainty.
It is a disciplined way to make judgment visible, explainable, debatable, and revisable.
What the Scoreboard Measures
Each Signal is read through the core question of The Race:
Does this development concentrate power or disperse democratic capability?
A development leans toward concentration when it strengthens already-powerful actors, deepens dependence, reduces accountability, limits meaningful access, or weakens democratic agency.
A development leans toward dispersion when it spreads useful capability, strengthens public or civic capacity, improves accountability, expands access, or gives more people and institutions meaningful power to act.
Many developments are mixed.
That is the point of the Scoreboard. It does not force every event into a simple good-or-bad category. It asks what direction the event appears to push, based on the evidence available at the time.
How Signals Are Scored
Each Signal receives a simple directional reading:
| Reading | Meaning |
|---|---|
| Strong concentration | Clearly strengthens concentrated power. |
| Concentration-leaning | Mostly strengthens concentrated power, though some effects may be mixed. |
| Mixed or unclear | Effects are balanced, uncertain, early, or too complex to classify cleanly. |
| Dispersion-leaning | Mostly spreads capability, accountability, or agency, though some effects may be mixed. |
| Strong dispersion | Clearly expands democratic capability or public/civic power. |
The reading is based on a small set of recurring questions:
- Power: Who gains leverage?
- Capacity: Who gains the ability to act?
- Accountability: Can the decision or system be understood, challenged, corrected, or governed?
- Access: Does meaningful AI capability become more broadly available?
- Agency: Do people and institutions gain more voice, choice, control, or participation?
The Scoreboard may sometimes use numbers internally to keep judgments consistent. But the public-facing score is the directional reading and the explanation behind it.
The explanation matters more than the label.
Confidence
Each Signal also receives a confidence rating.
| Confidence | Meaning |
|---|---|
| Low | The facts are limited, early, incomplete, or highly uncertain. |
| Medium | There is enough information for a reasonable provisional judgment, but important details could change the interpretation. |
| High | The evidence is strong, the implementation details are clear, and the interpretation is relatively stable. |
Confidence is not the same as intensity.
A development can be a strong concentration signal but still receive low confidence if the facts are incomplete. A modest dispersion signal can receive high confidence if the evidence is clear.
Current Readings and Longer Patterns
The Scoreboard will include two kinds of judgment.
First, it will provide a current reading: a plain-language summary of where the most recent signals appear to point.
Second, it will provide an interpretation: what those signals suggest about the broader race between concentration and dispersion.
Over time, the Scoreboard may include weekly, monthly, and quarterly updates.
- Weekly updates will capture recent movement.
- Monthly updates will look for patterns across multiple Signals.
- Quarterly updates will step back and ask whether the direction of travel is changing.
No single Signal decides the race.
The pattern matters.
Why Scores Can Change
Scores are provisional.
That is not a weakness. It is part of the method.
AI developments often look different after implementation than they do at announcement. A policy may sound accountable but fail in practice. A private tool may appear concentration-biased at first but later become a dispersion vehicle. A local fight may begin as opposition to infrastructure and become a model for democratic negotiation.
As more information becomes available, a Signal may be revised.
The goal is not to be frozen in the first judgment.
The goal is to get better over time.
A Developing Method
This scoring method is itself part of the research.
It will change as The Race develops, as more Signals are scored, as readers offer feedback, and as the project learns what produces the clearest and most useful public judgment.
The aim is to build a scoring system that does three things at once:
- Reflects reality without pretending complex events are simple.
- Produces meaningful comparisons across different kinds of AI developments.
- Stays easy enough to administer regularly and transparently.
The Scoreboard is not final.
It is a working public instrument.
Its purpose is to help readers see the race more clearly.