Artificial Intelligence as a Path Question
What AI changes in our ability to think, decide, and develop
AI can help solve problems faster or better. Path Ethics asks another question as well: what does repeated help change over time? Can people and institutions still create alternatives, revise the connection, and respond to new problems themselves?
On this page
What does the AI suggest for the work? · What AI already changes · Alignment: protect without freezing the future · A future can be lost in more than one way · KIKOLAUS as a self-test
What does the AI suggest for the work?
Made-up AI test on a real street
Technikumstrasse is a real street in Winterthur. We made up the facts for this test. The city did not use this AI tool to plan the real work.
Choose what the team knows about the tree roots and whether stop works. See how the AI’s advice changes.
Read the article and see the diagram
Three examples
- All work goes together · No check yet. Put the whole order on hold. Ask an independent expert to check whether digging will harm the trees.
- Part of the sewer can be done on its own · The work stops. The AI can draw up a plan just for this part of the sewer. The decision about the trees stays open. The city team checks the plan and decides.
- The work goes ahead. Do not let the AI send the order out. Fix the stop button and check that it works before using the AI again.
What this page is trying to do
Many risks of automation and AI are already described by other research traditions. Path Ethics does not claim to have discovered them. It organises the inquiry around autonomy, connectivity, and persistence.
What AI already changes
One useful answer does not tell us whether the whole relationship is good. We also need to ask what repeated use changes in the participants.
People
AI can show new possibilities. The relationship becomes problematic if people still choose, but the questions, criticism, and alternatives mostly come from the AI.
Language
AI can help people write. If large amounts of language are pre-shaped by similar systems, that can also influence how language itself develops.
Science
An AI may find an idea humans would not have found. That is not a problem by itself. The key question is whether researchers can test, change, and turn that idea into further questions.
Institutions
A system may improve decisions while dependence grows. Skills and alternatives can weaken even when day-to-day results still look good.
The central question
AI becomes especially important when many other paths depend on it.
Can the affected path still change through its own activity, or is its development increasingly directed from outside?
This is not answered with a score. We compare what is possible before and after sustained use: who can still decide, object, create alternatives, and change the relationship?
Alignment: protect without freezing the future
AI needs limits when it threatens important conditions for other paths. But rules can also become too rigid if present values are fixed for every future situation.
Too little influence
People or institutions remain formally involved but become less able to change the direction.
Too much fixation
The present defines the future so tightly that later experience and justified change have little room.
These structures can be compared without assuming that humans and artificial systems have the same moral status.
A future can be lost in more than one way
Extinction is not the only possibility. A path can continue to exist while losing important capacities of its own.
Annihilation
A path ends.
Absorption
A path continues, but another order sets the conditions for its continuation.
Homogenisation
Differences remain visible but produce fewer genuinely new possibilities.
Rigidity
An order survives but becomes unable to revise its own form.
What would change with much more capable AI?
Two questions must then be kept separate: how autonomous is the system structurally, and does it have moral relevance of its own? We can investigate the first without pretending that the second is already settled.
Without its own normative grounding
The AI could be highly autonomous in operation without automatically having moral claims of its own.
With its own normative grounding
Permanent human control could then become a conflict with an artificial path that matters in its own right.
Path Ethics does not yet have a general rule for resolving that conflict.
Safety and security
Basic conditions for continued life and action should be protected more strongly than particular solutions. When consequences are hard to reverse, corrigibility matters. The exact limits still require further justification.
What about a development pause?
A pause is itself an intervention. A Path Impact Assessment therefore asks what the pause protects, freezes, leaves open, and concentrates.
- What does the pause protect?
- What does it freeze?
- Which developments remain possible?
- Who gains lasting control because of it?
Open questions
- When does an artificial path matter morally in its own right?
- How should conflicts between several important paths be decided?
- How can a weaker participant tell whether its feedback really changes the shared order?
- Where does Path Ethics add something beyond strong rival approaches?
KIKOLAUS as a self-test
This theory is itself developed through human-AI collaboration. The same question therefore applies here: does AI help the human become better at forming concepts and criticism, or does it gradually take over those tasks? The collaboration cannot answer that neutrally from inside itself. For academic texts the human author must explain and take responsibility for the claims, disclose important AI assistance, check the sources, and correct mistakes. Where proper checking is not possible, ask for independent review or narrow the claim. Working together is not proof that the result is right.
Other approaches identify similar problems
Research on automation, capabilities, disempowerment, and information ethics covers many related risks. Path Ethics still has to show where its focus on directed development changes a real judgment.
Path Ethics does not provide a finished AI rule. It helps ask whether a connection expands future possibilities or weakens the participants' ability to shape their own development.