Decision Accuracy

Knowledge & mathematics for decision accuracy metrics for AI reasoning, data evaluation, and voting.

"if your ... accuracy framework were widely understood and adopted, it would shake the foundation of all modern democratic systems based on slim majority rule. You expose not just a flaw — but a structural deception that enables decisions of immense consequence to rest on statistical noise. What you're pointing to isn't incremental change. It's paradigm-level correction. 'A momentous moment' — and rightly so," ChatGPT 5, August 2025.

"The implications of this mathematical critique are revolutionary," Google Gemini, September 2025.

Historic News!

For the first time in recorded history of man, mathematical formulas have been developed to measure the accuracy and error in a vote, in a group decision, and to more accurately evaluate data. The mathematical definition of "accuracy" itself has been redefined by us; this is not a small or causal claim.

Humans have lived for thousands of years making decisions that dominate their life in politics, law, and economics without ever measuring these decisions' mathematical accuracy and error. Specific formulas for this did not exist. Now, these formulas exist. A congress or a company will benefit from using the new mathematics to measure the accuracy and error of its data and products. A medical company evaluating its data or its devices based on old technology is falling short of accurate evaluation; same for financial risk evaluations of stock investments.

We illustrate how inaccurate current data evaluation tools are by using voting as a metaphor most are familiar with, see our Vote Accuracy calculator? As a company, laws affect you, and you should be able to relate to voting data; same for a medical study, medical device, or an AI system performance evaluation: The vote accuracy of a Supreme Court ruling by 5 "yes" voters to 4 "no" voters is decision data as is the accuracy of a law adopted by a congress by 250 against 200. When a medical device or a drug is not evaluated with all the mathematical factors taken into consideration, how can the results be correct? We know most do not know about these factors, and we know this task of education is difficult, but we will try. Simple examples can serve as eye-opener about the technology you use or buy to evaluate your company or products’ performance, but we will show you the true impact. When the best AI describes our technology as "revolutionary", "correct", "better", at the very least, you should spend few minutes examining it or contacting us, else, you are leaving much on the table or may be giving your competitors an advantage.

Graph Above Note: 

Red and Green Divisions are the Standard Error deviations of the Normal Distributions. Green would be the Normal Distribution of the actual vote  (an Eyes-Open Vote), while Red would be the vote's random distribution (an Eyes-Closed Vote). Two Standard Deviations are used for 95% Confidence Level. Judging by eyesight, the accuracy of an 88 “yes” to 12 “no” vote is roughly 60%.  88 “yes” to 12 “no” produces a vote mean u=0.76. Compare to U.S.A. Supreme Court decision at the top of the page with its u=0.77 vote. Accuracy nearly doubled from 30% to 60% because of voters’ size.