26 October 2026
AI is making existing knowledge easier to access and synthesise. The harder task is working out what that knowledge means here, what other people see that we do not, and what happens when we act on it. Reflection may therefore become more important, not less, as AI becomes more capable.
Learning something unfamiliar used to involve quite a bit of searching. One had to find the right material, work out what mattered and gradually connect one idea with another. AI can now do some of that initial work remarkably quickly. Its answers still need to be questioned, but it has made it much easier to reach an informed starting point on a subject.
Making knowledge easier to access does not make knowing less important. If one knows very little about a subject, it is harder to recognise when an AI response is superficial, misleading or wrong. Prior knowledge helps us question what we are given, connect it with other ideas and notice when something does not quite add up. There is also a social reason for knowing things. Without some shared knowledge, even disagreement becomes difficult. We need enough in common to know what we are disagreeing about.
If AI becomes increasingly good at bringing together what is already known, what should learners be developing beyond the ability to obtain and repeat it? Judgement seems increasingly important: knowing how to use what is already known without assuming that it can tell us exactly what to do here. That means negotiating the difference between general knowledge and particular circumstances, between our own reading of a situation and what others see, and between an idea that sounds convincing and one that holds up when acted upon.
Some understanding only emerges when general ideas meet particular circumstances. A leadership theory can describe conflict, but it cannot determine exactly what is happening between five people in a particular room at a particular moment. The people involved bring histories, relationships, assumptions, emotions, cultural experiences and memories that affect how the situation is understood. Someone may notice a silence. Another person may interpret it as agreement. A third may experience it as fear. A shared framework may help them make sense of what is happening, but it cannot settle what that silence means.
Emotion is part of the situation too. It can interfere with judgement, but it can also draw attention to something that has not yet been properly understood. Feeling uneasy about a decision does not prove that the decision is wrong, but the feeling still gives us something to examine. If we are trying to understand an experience, emotion cannot simply be removed from it.
This is where reflection becomes particularly important in an AI-rich world. AI can quickly bring established explanations, theories and research to bear on a problem. Reflection allows us to ask how well that knowledge fits what we have actually experienced. What happened here? What did I notice? What did I feel, and why? What did somebody else see that I did not? Where does the existing explanation help, and where does it fall short?
Reflection still needs something to think with. Theories, models and research give us ways of interpreting experience and asking questions that might not otherwise occur to us. We do not question from nowhere; we question with tools that have been handed to us. But those same tools can begin to settle the meaning of an experience rather than helping us examine it.
Reflection does not make our interpretation reliable simply because we have thought deeply about it. People misread situations. Memory is fallible. Cultural assumptions may contain hard-won wisdom while also preserving prejudice or error. Being part of an experience gives us access to it, but not an unquestionable account of what it meant.
“My experience” cannot become the final answer to “what is true”.
If everyone retreats into the authority of their own experience, we face a kind of epistemic Tower of Babel: different accounts of reality with too little common ground between them. The problem is not that people see things differently. That difference can be valuable. The problem begins when there is no shared basis on which those different interpretations can meet, challenge one another or be revised.
This is where reflection has to become more than an individual exercise. Another person may have noticed something we missed, interpreted the same event differently, or brought experience that changes the questions being asked. Reflection becomes richer when our interpretation can be interrupted by somebody else’s.
But that requires some common ground. We need enough shared knowledge to think together without requiring everyone to think the same way.
The harder question is whose knowledge becomes part of that common ground. Some ideas come to be treated not simply as shared knowledge, but as universal. How did they acquire that status?
Some claims become widely accepted because they have survived systematic investigation, repeated challenge and comparison with competing explanations. That matters, and criticising the history of knowledge does not require us to treat every claim as equally well supported.
But knowledge also has a history. What gets recorded, institutionalised, published, taught and treated as authoritative is shaped partly by the distribution of resources, recognition and power. Different cultures have developed different ways of understanding relationships, community, nature, responsibility and what it means for a life to go well. Treating one tradition as though it came into being without a history can make other forms of understanding difficult to see.
AI sharpens this problem. The material available to these systems does not represent every culture, language or intellectual tradition equally. What has been recorded, digitised and made available shapes what a system can draw upon. An answer can therefore sound universal while drawing from a record in which some traditions are much better represented than others. Perspectives outside the dominant record matter for precisely this reason.
The answer is not simply to reverse the hierarchy. Knowledge does not become true because it has been marginalised, any more than it becomes false because it has been institutionalised. Nor does it help to reduce enormous and internally contested traditions to convenient categories such as “Western rationality”, “Eastern harmony” or “Indigenous wisdom”.
Perhaps the more useful question is what knowledge we are prepared to hold in common, rather than what deserves to be called universal. Shared knowledge does not need to be beyond challenge. It needs to be reliable enough for us to think and act together, while remaining open to evidence, experience and circumstances that expose its limits.
That still leaves difficult questions. What supports a claim? How was that support produced? What assumptions shaped the enquiry? What experiences were excluded? Does it continue to explain what we encounter, including circumstances different from those in which it began? These questions offer no infallible test, but they give us a basis for challenge. Institutional authority cannot make a claim unquestionable, just as personal or cultural authenticity cannot place one beyond scrutiny.
Reflection also has to remain open to what happens when an interpretation is acted upon. We can think carefully, draw on established knowledge and listen to other perspectives, and still be wrong. What happens next gives us something further to reflect on.
This is easier to see in early human communities. Knowledge about where to hunt, what to eat, what to avoid or how to respond to changes in the environment was developed and passed on within relatively small communities living in particular places. Much of that knowledge remained close to the conditions in which it was used. If something did not work, the consequences could also be experienced within the community.
That does not mean early communities always got things right. Beliefs could persist despite being wrong. The important point is the proximity: the people sharing knowledge, the people acting on it and at least some of the consequences belonged to the same social and physical world.
The internet changes that proximity. Knowledge and opinion can now travel almost completely free of the circumstances in which they arose. An interpretation of work, relationships, success, leadership or society can reach millions of people living in circumstances the person offering it may never encounter.
The grandfather telling stories by moonlight has, in one sense, acquired a global audience. The difference is that the new storyteller may share neither the community nor the environment of the people listening. Nor will the storyteller necessarily encounter what happens when those listeners begin to organise their lives around what they have heard.
This does not make the grandfather wise and the influencer foolish. Either can be wrong. The more interesting question is what happens when they are wrong. What allows the idea to be challenged? Who experiences the consequences? Can those consequences find their way back to the person promoting the idea? And is there anything that requires that person to reconsider what they said?
The same question arises when AI informs a decision. Consider an AI-supported recommendation to restructure a healthcare service. The recommendation may look convincing, and later data may show changes in waiting times, expenditure, staff turnover or patient outcomes. But those outcomes should not simply become another set of data points. They are part of what happened, and therefore part of what has to be reflected upon.
Staff live through the restructuring. Patients experience changes in care. An improvement in waiting times might sit alongside a deterioration somewhere else. What follows from the decision can therefore change how the original recommendation is understood.
Acting on an interpretation produces new experience, and that experience may give us reasons to confirm, question or revise what we thought we knew. AI can help us analyse what happened, but it does not remove the need for people to reconsider their judgement in light of what followed.
There is a distinction here between consequence and answerability. Consequences may give us reasons to revisit a judgement. Answerability concerns who remains responsible for making that judgement and acting on it. AI may inform a decision, but it can also make it easier to hide behind a model or recommendation. Reflection therefore matters not only because we may have been wrong, but because people remain answerable for the judgements they make.
By this point, reflection has become more demanding than simply thinking back on an experience. It means bringing what happened into conversation with what is already known, allowing other people to challenge our interpretation, and returning to our judgement when we see what follows from acting on it.
Think again about the silence in the room. A theory of conflict might help explain what is happening, but it cannot settle what that silence means. One person reads agreement; another experiences fear. Neither interpretation becomes true simply because someone experienced it that way. Understanding develops as those different readings are considered alongside what is already known and remain open to what happens next.
AI can contribute to this. It can bring forward relevant theories, offer other interpretations, organise observations and help us consider possibilities we may have missed. But none of that settles what the situation means or what should be done. Judgement is still required. Someone has to decide what matters here, what does not fit, whose experience may be missing, and what should follow.
This is where reflection becomes judgement. We cannot keep reflecting indefinitely. At some point, something has to be decided or done, often without the certainty that we would like to have. Reflection does not remove that uncertainty. It helps us understand what our judgement rests on and where its limits might be.
A reflective judgement might therefore sound something like this:
This is what I noticed. This is what I know. This is where I may be wrong. This is how my experience has shaped what I see. This is what established knowledge suggests. This is what somebody else sees differently. And this is the judgement I am prepared to make for now.
“For now” matters. It does not make the judgement tentative to the point of uselessness, nor does it mean that anything goes. It recognises that a decision sometimes has to be made before certainty is available. What happens next may confirm the judgement, complicate it or give us reason to think again.
If access to accumulated knowledge becomes abundant, one increasingly important purpose of learning is to develop people capable of exercising judgement when knowledge alone cannot determine what should be done.
This does not make acquiring knowledge obsolete. Judgement without knowledge is poorly equipped to distinguish insight from impulse. But knowledge without judgement can leave us with sophisticated explanations and little capacity to decide what those explanations require of us.
We need shared knowledge to think together, and sufficient freedom from it to think differently. What is already known has to meet what is happening here. Our interpretation has to meet the interpretations of others. And what we decide has to remain open to what happens when we act on it.
Reflection helps keep these things in conversation.
If that is right, AI makes what happens after the answer more important, not less. Increasingly persuasive answers do not remove the need to work out what they mean in a particular situation. We still have to recognise when an answer does not fit, decide what should follow, reflect on what happens when we act, and remain answerable for the judgement we make.