The Data Is Lying to You
And Nobody Is Responsible For It
Every consequential decision made in the world today rests on data. Which drugs get approved. Which neighbourhoods get policed. Which communities receive climate investment. Which voices get amplified by algorithms and which get buried. Whose children inherit a world shaped by models trained on records of the past.
We have built civilisation's decision-making infrastructure on a foundation nobody has examined.
The Problem Nobody Names
When something goes wrong with an AI system, we investigate the model. When a policy fails, we audit the process. When a scientific finding turns out to be false, we blame the researchers.
We almost never ask the prior question: who decided what data to collect, under what conditions, in whose interests, and what did they leave out?
Data does not fall from the sky. Every dataset is the product of decisions — about what to measure and what to ignore, which populations to include and which to exclude, what counts as a valid finding and what gets filed away. These decisions are made by human beings operating inside institutions that have interests, funding relationships, and incentive structures that shape what they choose to see.
A pharmaceutical company funds a clinical trial. The scope — which outcomes to measure, over which timeframe, in which populations — is set before a single patient is enrolled. A scientist working honestly within that scope produces honest findings. But the scope itself was never neutral. It was drawn to answer the questions the funder needed answered, not all the questions worth asking.
The finding that would have undermined the product never gets measured. The population that would have shown a different result is not in the study. The anomaly that appeared in week three gets noted and not followed up.
None of this requires fraud. It requires only that the people who fund knowledge production have interests — and that the people who produce knowledge need funding.
This is not an edge case. This is the default condition of data production in the modern world.
Why Everything We've Built to Fix This Misses the Point
The problem is not for lack of effort. Frameworks have been built, regulations written, research centres funded, expert panels convened, guidelines published.
They share one fatal assumption: that the data already exists, and the ethical work begins downstream of it.
Data protection law tells us who can access the data. It does not ask how the data was constituted or whose interests shaped its production.
Algorithmic accountability frameworks audit what AI systems do with the data. They cannot detect distortions that entered before the model ever saw a single record.
Open data initiatives make the data available to everyone. Democratising access to a distorted dataset does not correct the distortion — it universalises it.
Fairness-aware machine learning measures whether a model performs equitably across groups represented in the evaluation set. It cannot measure the impact of groups who were never in the dataset at all.
Each of these interventions is genuine. Each addresses something real. And each takes the data as a given — as a pre-ethical fact about the world — rather than as what it actually is: a social product constituted inside human institutions that carry human interests.
The result is a governance landscape that produces false confidence. Organisations can be fully compliant with every data regulation, fully audited for algorithmic fairness, fully committed to open access — and still be making decisions on a foundation of systematically distorted knowledge. The scrutiny was applied. It just was never applied to the layer where the problem lives.
What Accountability Actually Requires
If data is constituted rather than discovered — if every dataset reflects the decisions of the people who produced it — then accountability requires making those decisions visible.
Who funded this data's production, and what were their interests? What was measured, and what was deliberately or structurally excluded? What assumptions were built into the methodology? What would the findings look like if different choices had been made? Who was affected by the conclusions, and did anyone ask them?
These are not exotic questions. They are the questions any serious empirical inquiry demands. They are simply not the questions that any existing governance framework requires to be answered.
The OpenReason Protocol exists to change that.
The Protocol
ORP is an open standard for transparent, reproducible, and accountable reasoning built on data. It does not tell you what conclusions to reach. It creates the conditions under which the reasoning toward any conclusion can be examined.
It works in five layers.
Data Provenance — every dataset documented at its source: methodology, funding, exclusions, limitations, and the decisions that shaped what was collected and what was not.
Consequence Simulation — explicit modelling of what the data implies under multiple scenarios, with assumptions stated and uncertainties acknowledged, including the downstream effects of the exclusions documented in Layer 1.
Empathy Mapping — identification of every party affected by the reasoning, including those absent from the data, with documented analysis of how the conclusions affect them.
Accountability Ledger — a traceable record of every significant decision in the reasoning process: who made it, when, and why, with alternatives considered.
Fork Registry — a formal mechanism for proposing alternative assumptions and comparing what different starting points produce, so that the relationship between choices and conclusions is permanently visible.
ORP is not a regulation. It is not a compliance checklist. It is infrastructure — the epistemic equivalent of what double-entry bookkeeping did for financial accountability. It makes the reasoning visible so that the reasoning can be challenged.
Why Now
Artificial intelligence is not a technology problem. It is an epistemological problem. The systems being deployed at civilisational scale are built on data produced under conditions nobody is examining, reflecting interests nobody is disclosing, encoding exclusions nobody is documenting.
We are automating the distortions of the past and calling it progress.
The moment to address this is before those systems become the infrastructure of governance, medicine, education, and economic life. That moment is now — not because the technology is new, but because the window to establish accountability at the foundational layer is closing.
Every year that passes without epistemic standards for data constitution is a year in which the systems being built inherit undocumented distortions that will be progressively harder to trace and correct.
ORP is open. It is free. It is already working.
The question is not whether we can afford to adopt it.
It is whether we can afford not to.
— Asbjørn Rasmussen, founder
The OpenReason Protocol is an open standard developed by the Public Reason Project.
Documentation, specification, and tools: openreasonprotocol.org
Contact: hello@publicreasonproject.org