One of the more interesting questions about AI is not whether it can solve problems. It often can. In some domains, it can solve them remarkably well. The more important question is whether it can get better at framing problems, not just answering them. That distinction matters more than many people realize.
Problem solving begins after the problem has already been defined. Problem framing comes earlier. It asks what the real issue is, what assumptions are hidden inside the question, what has been left out, and what kind of answer would even count as meaningful. Many failures in business, policy, education, and research do not happen because people were unable to solve a problem. They happen because the wrong problem was being solved in the first place. This is where current AI often shows both its strength and its limits.
Give AI a sufficiently clear task and it can generate options, compare tradeoffs, optimize variables, summarize positions, and produce elegant responses. But ask it whether the task itself has been framed well, and its performance becomes more uneven. It tends to inherit the framing it is given. It may challenge it, but often not with the same consistency or depth that a strong human thinker can bring.
That raises an intriguing possibility: can AI become better at problem framing if it is guided into different cognitive stances? More simply, would it help if AI tried to think more like Einstein, Wilber, or some other distinctive thinker? I think the answer is yes, up to a point.
The important thing is to understand what “thinking like” someone means in this context. It does not mean literal imitation. AI does not become Einstein. It does not somehow access Wilber’s mind. Nor does it become any specific user in a deep sense. What it can do is approximate a mode of inquiry. It can adopt a style of questioning, a pattern of attention, a way of organizing complexity, or a preferred form of synthesis. And that can matter a great deal.
Take Einstein as an example. An Einstein-like stance would push AI toward simplification, first principles, and thought experiments. It would encourage the model to distrust surface complexity and search for the underlying structure beneath appearances. Instead of merely responding to a crowded set of variables, it might ask whether some of those variables are secondary rather than fundamental. It might seek the governing dynamic rather than the local symptom. That can be extremely useful in reframing messy situations.
Now consider Ken Wilber, the founder of Integral Theory. A Wilber-like stance would lead AI toward multi-perspectival integration. It would be more likely to look across levels, domains, and dimensions. It might ask whether a problem has interior and exterior aspects, individual and collective aspects, developmental aspects, and cultural aspects. It would resist reductionism and try to situate a problem in a larger map. That can be useful when dealing with complexity that cannot be adequately captured by a single lens.
Then there is a third possibility that may be even more relevant for advanced inquiry: asking AI to think in a more architectonic or metatheoretical way. In this style, the task is not merely to answer a question or even integrate perspectives. It is to examine the assumptions shaping the inquiry itself. What ontology is implicit here? What epistemology is being privileged? What values are embedded in the framing? What telos is guiding the process? What practical orientation is assumed? In such a mode, AI becomes less of a responder and more of a reflective partner in interrogating the frame.
This matters because much of current AI use is still too generic. A user asks a question and the model produces a plausible answer in an averaged, middle-of-the-road voice. That is often useful. But it is also one of the reasons AI can remain trapped inside conventional framings. A generic intelligence tends to reproduce generic assumptions. If you want better framing, one way to get it is to move the system out of its default cognitive center of gravity.
In that sense, asking AI to “think like” a distinctive thinker can function as a practical intervention against intellectual mediocrity. It can also help compensate for one of AI’s most important weaknesses: premature closure. Large language models are often too quick to converge on the first coherent answer. They can appear reflective while still accepting the basic terms of the question too readily. A strong cognitive stance can disrupt that tendency. It can force the model to ask different questions before attempting to answer. Not always, but often enough to matter.
Still, this approach has real limits. Every cognitive stance has its biases. Einstein-like thinking may overvalue elegance and simplicity, even when the problem is socially messy and irreducibly plural. Wilber-like thinking may over-integrate and impose a grand map where more fragmentation or ambiguity should be preserved. A metatheoretical stance may reveal hidden assumptions beautifully while underplaying practical constraints on the ground. No perspective is neutral. Each one sharpens some things and distorts others. So the goal is not to find the one best thinker for AI to imitate. The goal is to use multiple stances as lenses.
This is where the method becomes more powerful. Instead of asking AI to produce a single answer, one could ask it to move through a sequence of reframings. Start with a default analysis. Then ask for an Einstein-like reformulation focused on first principles. Then a Wilber-like integration of multiple dimensions. Then a metatheoretical critique of underlying assumptions, boundaries, and purposes. Finally, ask for a synthesis across all of them. That process is far more likely to produce genuine reframing than a single-pass response.
The broader implication is that better problem framing in AI may depend less on raw scale and more on cognitive orchestration. Bigger models will help in some respects, but raw intelligence alone does not guarantee better framing. Framing requires disciplined perspective-taking, boundary critique, normative sensitivity, and the ability to consider alternative definitions of the problem before rushing toward solutions.
That is why I suspect the future of AI framing will not come mainly from isolated model improvement. It will come from systems and practices that structure reflection more deliberately. Some of that structure may be built into the interface. Some may come from multi-agent architectures where different models play different epistemic roles. Some may come from users who know how to push the model beyond generic answer production. And some may come from new benchmarks that reward reframing rather than merely correct answer generation. In that sense, asking AI to think more like Einstein, Wilber, or any other strong thinker is not a gimmick. It is an early, imperfect version of something deeper: teaching AI to move among intellectual postures rather than remaining trapped in one.
There is also a human lesson here. The value of AI increasingly depends on the quality of mind that engages it. If the user accepts the first framing that appears, AI may simply reinforce shallow thinking. But if the user knows how to introduce alternative stances, competing frames, and deeper questions, the system becomes much more generative. The difference is not only in the model. It is in the interaction.
This is one reason I am skeptical when people talk about AI as though it were simply a mirror. It is not merely reflecting the user back to themselves. It is better understood as a responsive medium whose outputs depend heavily on the form of inquiry brought to it. The prompt is not just an instruction. It is an epistemic intervention. That means “think like Einstein” is not really about Einstein. It is shorthand for: simplify more radically, seek the hidden principle, question surface assumptions. “Think like Wilber” is not really about Wilber. It means: widen the frame, include more dimensions, resist flattening. And asking AI to think in a more architectonic or metatheoretical way means: do not take the question at face value. Interrogate the structure behind it.
So yes, AI can become better at framing when guided by strong cognitive lenses. But the real breakthrough will come when those lenses are not occasional tricks but part of how AI is routinely designed and used. Until then, one of the best ways to get better framing out of AI may be surprisingly simple: stop asking it only for answers, and start asking it what kind of mind should approach the question in the first place.
AI may not need to become Einstein or Wilber. But it may need to learn how to borrow better minds than its default one