Skip to content
Advancing Bits to Master Atoms

Learning cause and effect.
Your AI learns which action led to which outcome.

MS Duplex keeps the link between what was done, the conditions it was done under and what happened next. It uses those links to recommend the next step, and tells your researchers when a link has not yet been established.

Learn and improve · one of the fourteen capabilities of MS Duplex

A row of satin-silver spheres hanging in a graphite frame, one cobalt-blue sphere swung out at the end, about to strike the row.A row of satin-silver spheres hanging in a graphite frame, one cobalt-blue sphere swung out at the end, about to strike the row.

Your team gets

  • Recommendations grounded in your own outcomes
  • Clear separation of what is known from what is assumed
  • Better choices of the next experiment or run

What learning cause and effect means for your team.

Learning cause and effect means MS Duplex keeps more than the result. For each confirmed outcome, it keeps what was done, the conditions it was done under and what happened next, linked together, so your AI can tell which action led to which outcome.

Those links shape what your agents recommend. When your team asks for the next step, MS Duplex draws on the actions that led to the outcome you want, under conditions like the current ones, and sets aside the actions that led elsewhere.

It also says what it does not yet know. Where a link between an action and an outcome has not been established, MS Duplex tells your researchers, so your team can tell a finding from a hunch and choose the next experiment with that in view.

Why research needs AI that knows why.

Two results can look alike and have different causes. An AI that keeps only the result cannot tell which action to repeat.

In research, a result matters because of how it came about. A run succeeded because of a setting, under a certain condition, after a certain step. Change the condition and the same action may lead somewhere else. The decision about what to try next depends on knowing which part of the work made the difference.

When an AI keeps only outcomes, it repeats whatever happened to come before a success and treats a coincidence as a rule. Its recommendations sound equally sure whether they rest on many confirmed results or on one lucky run, and your researchers have to work out for themselves which is which.

  • Conditions decide

    An action that works under one set of conditions can fail under another, so the conditions have to travel with the link.

  • Findings and hunches differ

    A team needs to know which recommendations rest on established links and which are still assumptions.

  • The next run is expensive

    Choosing the next experiment well depends on knowing which actions moved the outcome.

  • Decisions get questioned

    A recommendation that shows the action, the conditions and the outcome behind it can be checked by anyone.

A branching line of upright graphite tiles on a dark plinth, a cobalt-blue tile at the start and one path of fallen silver tiles leading into a single branch.A branching line of upright graphite tiles on a dark plinth, a cobalt-blue tile at the start and one path of fallen silver tiles leading into a single branch.

How learning cause and effect works in your team’s day.

Your team records what it did and what happened, as it does today. MS Duplex keeps the links and uses them for the next recommendation.

  1. 1

    An action is taken

    The agent or a researcher runs a step with chosen settings, under known conditions.

  2. 2

    The outcome is confirmed

    A test, a measurement or an expert confirms what happened next.

  3. 3

    The link is kept

    MS Duplex keeps the action, the conditions and the outcome together, with the evidence behind them.

  4. 4

    The next step is recommended

    Recommendations draw on established links for conditions like the current ones, and a link not yet established is labeled as such.

An example, in any field where an agent recommends the next run.

A team’s agent recommends settings for the next run toward a target. Its recommendations grow clearer about what is known and what is not.

  1. 1

    A pattern appears

    Several successful runs share two settings. Each run is recorded with its actions, its conditions and its confirmed outcome.

  2. 2

    A link is established

    Further confirmed runs show that one of the two settings leads to the target under the team’s usual conditions, and the other made no difference. MS Duplex keeps the first as an established link.

  3. 3

    A new condition

    A researcher asks for a recommendation under a condition the team has not tested. MS Duplex recommends the established setting and says that the link has not yet been established under this condition.

  4. 4

    The next experiment

    The team runs that case as a deliberate test. The confirmed result extends the link to the new condition or marks where it stops holding, and the next recommendation reflects it.

Cause and effect gives the other capabilities their reasons.

Knowing why an outcome happened makes every plan, decision and improvement easier to trust.

  • Simulation before action

    Before a costly step, the options are played forward using the links your team has confirmed, so the budget goes where the evidence points.

  • Decisions that know their limits

    Where an established link clearly applies, MS Duplex decides directly; where it does not, the decision goes to your model or a researcher, with the reason.

  • Self-improvement

    Knowing which action caused which outcome lets your agents improve in the right place, trying first what led to success.

  • Provenance, review and unlearning

    Every link keeps the results it rests on, so another researcher can rebuild a recommendation and check it.

What your team gets.

Recommendations grounded in your own outcomes, with a clear line between what is known and what is assumed.

  • Recommendations from your outcomes

    The next step is recommended from what your own confirmed work shows led to the result you want.

  • Known apart from assumed

    Each recommendation shows whether the link behind it is established or not yet established.

  • Conditions kept with every link

    An action is recommended where the conditions match the ones it was confirmed under.

  • Better choices of the next run

    Your team can aim the next experiment or run at the links that matter most and are not yet established.

  • Reasons on record

    Every recommendation shows the action, the conditions and the outcome behind it, ready to review.

  • Links that carry

    The links your team establishes carry across sessions, agents and models.

Common questions.

How does MS Duplex tell a finding from a hunch?

It treats a link between an action and an outcome as established only as far as your confirmed results support it. Where a link has not yet been established, it says so in the recommendation, so your researchers can see which is which.

What happens when conditions change?

Each link is kept with the conditions it was confirmed under. When a case falls outside them, MS Duplex says that the link has not yet been established there, so your team can decide whether to test it.

Make the next run benefit from the last.

Bring us the research your AI should get better at. We will plan a pilot on your own workflows, with your experts as the judges.