Facts, sources, interpretations, assumptions, unknowns and alternative models — without collapsing them into one layer.
P-120 RESEARCH · RESEARCH · IN DEVELOPMENT
How we make decisions
A research project for a human-governed cognitive analysis system
More happens between information and a decision than it may seem. We see facts, interpret them, make assumptions, treat some things as obvious, and gradually arrive at a choice. Yet the path to the decision itself usually remains almost invisible.
We are investigating whether that path can be made observable and reviewable — not so that a system decides for a person, but so that a person can better see what their own decision rests on.
HOW THIS LOOKS IN PRACTICE
One decision. Several layers that are easy to confuse.
In a difficult decision, facts rarely stand alone. Explanations, expectations, fears and assumptions quickly accumulate around them — and after a while they can begin to feel equally certain.
HG-CGA helps separate these layers and identify what information is actually missing, without making the decision for the person.
01
RELATIONSHIPS AND A SHARED FUTURE
Should I keep building a shared future with this person?
We have been together for almost two years and have talked more than once about a shared future. My partner has repeatedly said they want us to live together, but moving in together has already been postponed twice. Most recently, they explained the delay by uncertainty around work and finances. I am beginning to think it may not be just circumstances: perhaps they are simply not ready to build a shared future with me.
- FACT
- The relationship has lasted almost two years. Conversations about a shared future have taken place. Moving in together was planned and has been postponed twice.
- SOURCE
- My partner says the latest postponement is connected to work and financial circumstances.
Important: the fact is that this explanation was given. That does not yet establish that the explanation is the true cause.
- INTERPRETATION
- “Repeated postponements may mean that my partner is avoiding a serious commitment or does not see our shared future in the same way I do.”
- ASSUMPTION
- “If someone genuinely wants a shared future, they should be ready to move in together now.”
- UNKNOWN
- Whether the obstacles are temporary. What conditions would need to change. Whether both people mean the same thing by “a shared future” and expect the same pace of progression.
- ALTERNATIVE MODEL
- The intention to build a shared future may be real, while readiness for the next concrete step is currently limited by circumstances, risk tolerance or a mismatch in pace.
This is not the system's conclusion about my partner's motives. It is only another model compatible with what is already known.
- CRITICAL QUESTION
- What specific action, condition or agreement within a mutually agreed timeframe would allow me to distinguish temporary unreadiness to live together from a lack of sustained intention to build a shared future?
What HG-CGA does here
It does not tell me to “stay” or “leave,” and it does not determine how sincere my partner is. It shows which parts of the decision are facts, which are interpretations, what assumptions connect them, and what new information could genuinely distinguish competing explanations.
Not every uncertainty is about another person's intentions.
Sometimes the main source of uncertainty is the future itself.
02
RELOCATION AND LIFE TRAJECTORY
Should I move to another country and change the life I know?
I have been offered a job in another country with higher pay and an opportunity for professional growth. I need to decide within the next few months. Here I have a stable life and people close to me. I am afraid both of missing a rare opportunity and of moving only to discover later that the new life does not suit me.
- FACT
- There is a concrete offer, a limited decision window, higher pay and a requirement to relocate.
- SOURCE
- The formal terms of the offer and the available information about the employer.
Important: stated professional growth is not guaranteed. Current stability also does not guarantee that life will remain unchanged.
- INTERPRETATION
- “If I turn this down, I may miss a rare opportunity.” At the same time: “If I move, I may lose what is already good and stable in my life.”
- ASSUMPTION
- “An opportunity like this will not come again.” “My current life will stay roughly the same.” “If the move is a mistake, it will be very difficult to reverse.”
- UNKNOWN
- Whether the new job will match expectations. Whether the new environment will suit me. Which parts of the decision are truly hard to reverse. Whether I can preserve important relationships and the option to return.
- ALTERNATIVE MODEL
- The choice may not be completely binary: some uncertainty may be reducible through additional information or by making part of the decision more reversible.
HG-CGA does not claim that such an option exists in every case or that it is the better choice.
- CRITICAL QUESTION
- What additional information, or what possibility of partial reversibility, would most change my willingness to move or stay?
What HG-CGA does here
It does not choose between “stay” and “move.” It separates actual conditions from forecasts about the future and shows what could genuinely change the decision.
The same structure remains when the decision becomes professional or operational.
03
OPERATIONAL DECISION · TRANSFERABILITY
The same architecture in a different domain
The delivery was nine days late. The supplier says it was a logistics problem, but I think the issue is their reliability.
- FACT
- The delivery was nine days late.
- SOURCE
- The supplier attributes the delay to logistics. The fact that this explanation was given does not establish that it is the true cause.
- INTERPRETATION
- “The supplier may be unreliable.”
- ASSUMPTION
- The cause of the delay was under the supplier's control.
- UNKNOWN
- Whether external logistics disruptions occurred that the supplier could not prevent.
- ALTERNATIVE MODEL
- The supplier may have acted correctly while the delay arose outside their control. This is a competing explanation, not an exoneration.
- CRITICAL QUESTION
- What information would distinguish a failure by the supplier from an external logistics cause?
HG-CGA does not decide whether to replace the supplier. It shows where the fact ends and the reliability inference begins — and what information is needed before making that decision.
DIFFERENT SITUATIONS · ONE COGNITIVE TASK
Different domains. The same structural problem.
Relationships, relocation and an operational decision look very different. But inside each one the same task reappears:
separate what was observed from how it is explained, expose hidden assumptions, keep unknowns unknown, and identify the question that could genuinely change the decision.
HG-CGA does not remove uncertainty or turn a difficult choice into an automatic answer. It makes the structure of the decision more explicit: what is known, where it comes from, what is already an interpretation, what assumptions are being used and what information is still missing.
04
FORMALIZED CAPABILITIES
What is already formalized in the architecture
Material contradictions, provenance gaps and unknowns remain visible rather than being silently promoted into established fact.
Decision history is preserved as an append-only sequence: new evidence may trigger review, but prior human determinations are not rewritten.
05
WHAT WE WANT TO TEST
Compare ways of reasoning on the same tasks
The central research question is not whether a computational system can “think better than a person.” We want to test whether the decision process can be made more transparent, reviewable and robust to particular classes of error without taking final judgment away from the person.
Ordinary free-form reasoning.
A strong structured approach to decision-making.
An experimental human-governed architecture.
06
WHERE WE ARE NOW
Not a percentage complete, but an evidence state
Formal architectureformalized
Executable governance layerimplemented and technically validated
Persistent decision historyreference persistence implemented and validated
Human pretestprepared · external ethics determination pending
Comparative studynot yet conducted
Empirical effectivenessnot established
The reference architecture has passed controlled technical validation.
PASS 9 established an executable governance layer. PASS 10 added event-sourced reference persistence and a replaceable multi-model adapter boundary. Decision history is append-only and replayable rather than rewritten.
Connected models remain sources of candidate outputs only: provider identity, model capability or agreement among several models does not by itself increase epistemic authority. Real external model-provider calls and PostgreSQL staging deployment were outside PASS 10.
Evidence boundary: these results establish technical conformance of the reference architecture to its governance constraints. They do not establish that HG-CGA improves human decisions.
The empirical effectiveness of the system has not been established.
HG-CGA is currently not a proven technology for improving decisions. It is a formalized research hypothesis and a programme for testing it. An external ethics-review route has been initiated.
P-120 RESEARCH
A faster answer is not always the point.
Sometimes it matters more to understand why this answer became ours.
Back to top