1. How we order experience
When we enter a room, we usually find our way around without first having to work out what is in front of us. We recognize a table, look for a free seat, and perhaps notice a tense atmosphere. Things appear in particular places, something happens before or after something else, an object seems familiar. For the moment, this understanding is enough: we have a rough sense of where we are and what we can do.
This very familiarity deserves a closer look. How do we recognize something as an object? What do we treat as belonging together, and what changes do we allow when we recognize something again? Two chairs can belong to the same kind, while we continue to identify an individual chair as the same one even when we see it from different angles. Being of the same kind and being the same individual thing allow different comparisons. We usually rely on such distinctions without considering them explicitly.
From this perspective, knowing can be understood as an activity that orders experience. Even what we treat as an individual object, an event, or a connection is part of this ordering. Kant asked what conditions make experience possible in the first place. The project takes up this question with regard to the ways we organize experience in terms of space, time, and objects. It leaves open whether any of these forms are necessary, which ones might be, and whether they can emerge or change under particular conditions. Ideas and expectations also play a part that we can barely put into words. A word such as “chair” brings some of this together without revealing all the relations that sustain our understanding.
The Epistemic Reality Project investigates this activity of ordering. It asks how we come to make sense of experience, how we maintain points of reference, and under what conditions our descriptions hold up. To do this, we need some distance from our habitual way of seeing things. We can examine what it enables and what it presupposes even if we have no alternative way of ordering experience. Such an inquiry may lead us to change our view, or to justify it more carefully and define its limits.
What we pick out also depends on what we intend to do. Someone looking for a place to sit attends to different features from someone preparing the room for an event. Our way of ordering experience helps us recognize possibilities and act on them. When we move or change something, we also learn whether our expectations hold up and which conditions we need to take into account.
These activities are also needed when science develops models or a society organizes shared procedures. There, they are elaborated through specialized methods and distributed across many participants. Everyday life offers an initial way to understand how we connect experiences, form expectations, and respond to deviations. An ordinary work situation provides a starting point: our assessment of a colleague we rely on for a shared task.
2. How a view comes to be taken for granted
A colleague with whom we regularly prepare events has carried out agreed tasks, met deadlines, and passed on important information. We consider her reliable and plan accordingly. This trust eases our workload: we do not have to reassess her reliability from scratch every time we divide up the tasks. She can rely on our work as well, so we can arrange many things together without having to agree on every step separately.
Then a question about the next event goes unanswered for two working days. At first, we think little of it; later, our plans become uncertain. Our colleague may be busy with other tasks or may have overlooked the message. But we may also have assumed a quick reply without having agreed on one. The missing reply brings to light an expectation that had previously guided our approach without our noticing.
This reveals a tension that runs through the whole project. We need a sufficiently settled understanding to find our bearings. Yet this very stability can make us overlook where our view falls short: as long as our expectations are met, we have little reason to question their assumptions. The unanswered message is only a minor disruption. But how we respond helps determine which experiences will count in our understanding from then on. We can overlook it, fundamentally change our view of our colleague, or first examine which expectation is actually affected.
The description “reliable” connects different experiences with an expectation about new situations. The project calls this process of sustaining continuity stabilization. It enables us to return to something and notice changes in it. Because we recognize our colleague as the same person across different situations and compare her behavior, we can trust her while also noticing when something has changed.
Yet the description contains more than we explicitly examine at any given moment. Someone who carries out agreed tasks thoroughly need not be available at short notice. In our general picture of reliability, however, these two expectations can merge. We then expect something of our colleague for which we have little evidence of our own.
The missing reply gives us reason to look more closely at this picture. What do we actually mean by reliability? Which experiences support it, and what expectation do we derive from them? Once we specify such relations sufficiently to compare, test, and change them, we are working with a model. It orders experiences so that expectations and possible next steps emerge from them.
Much of our everyday understanding remains less explicit. We sense that a situation is familiar or find our way through it without being able to name the relations at work. These forms of orientation need not already be fully developed models. They can still shape our behavior and only become the subject of closer inquiry when a difficulty arises.
The fact that we bring experiences together reflects our finite capacities. Attention, memory, and time are limited; we have to select which differences matter for our question. This selection is what makes certain connections manageable. A useful understanding therefore contains less than everything that could be said about a person or a situation.
We can lose sight of the benefit of this simplification, however, if we treat the resulting picture as complete knowledge. Then “I have been able to rely on her so far” turns almost imperceptibly into “That is how she is.” The conditions of our assessment disappear behind the quality we attribute to the person. We may stop distinguishing between situations, even though those very differences could explain why an expectation is now going unmet.
Even a poorly justified view can come to be taken for granted in this way. Repetition and familiarity make it persist, while experiences that fit become easier to notice than those that do not. The question of its validity therefore begins when we distinguish its mere stability from the reasons that actually support it.
3. What we call reality
By changing position, we can gain a different view of the room. What looked like a free seat turns out, on closer inspection, to be taken. When we move a chair, we learn how it can be moved and what stands in its way. Our intention directs our attention; the possibilities and limits we encounter can, in turn, change our view.
Distances, too, can be determined in different ways. In everyday life, something appears near, far away, or within reach. Through comparison and measurement, we can treat distances as quantities and express them in numbers. This lets us formulate mathematical relations and extend calculations to distances we can barely picture. Greater precision opens up new possibilities and also brings assumptions of its own: which quantities are measured and under what conditions their relations hold remain matters for investigation.
The theory of relativity shows how fundamentally familiar assumptions can change. In classical physics, the length of an object was taken to be independent of how an observer moves relative to it. Relativity shows, by contrast, that lengths and time intervals are determined differently in frames of reference moving relative to one another. A familiar basic assumption therefore had to be abandoned. Yet for many tasks at speeds far below the speed of light, classical descriptions remain sufficiently accurate approximations. Correcting a basic assumption can thus go hand in hand with continuing to use established descriptions within limits. In everyday life, we similarly ask which expectation about our colleague we need to change and which experiences we can continue to rely on.
Moving, intervening, and measuring also show us that the room does not conform to our expectations in every respect. Limits and resistance cannot be removed simply by changing a description. An understanding gains weight when it withstands such tests and allows correction. An unmet expectation, however, initially only gives us reason to investigate whether we need to reconsider our assumption, our approach, or how we characterize the situation.
Even our better descriptions remain tied to a way of gaining access. Another person’s account, a photograph, or a measurement must be taken in and understood. We can examine how our perception and concepts contribute, and extend their limits through further ways of gaining access. But a more accurate picture of the room is different from a view independent of every way of knowing.
The project speaks here of “Realism without an Outside.” We can describe real objects and test our statements about them. It does not follow that we can also give a complete account of what these objects are like independently of all conditions of knowing. Such a further claim needs additional reasons. Conversely, the fact that our descriptions depend on particular ways of gaining access does not settle whether the objects described exist independently of us.
What matters for finding our bearings is that we can justify descriptions, connect them, and correct them. We may treat the room as a real room and our colleague as a person in her own right whose behavior can surprise us. A well-supported way of gaining access achieves more than an arbitrary guess, without already giving us a final picture of reality.
4. How an assessment holds up
With our colleague, too, we need to distinguish more carefully what we want to know about the unanswered message. Our unease, the meaning of an agreement, and the expectation of a quick reply are connected. But they call for different approaches and reasons.
Our unease is initially accessible through our own experience. We can try to understand it more precisely: are we worried about the deadline, annoyed about extra work, or mainly missing the clarity we need to plan? This clarification can change our first interpretation. We may already have read the lack of a reply as a lack of interest in the shared task before asking ourselves what is bothering us about it.
To understand what was agreed, by contrast, we need to consider the agreement itself. Was a quick reply explicitly promised? Did both of us understand the same thing by that? A message can record the wording, but what that wording meant between us also depends on the conversation and our shared expectations. Our colleague can add to our recollection or dispute our interpretation.
Finally, to find out whether past behavior supports the expectation of a quick reply, we can compare relevant situations. What matters is which situations are comparable. A quick reply on a quiet working day may tell us little about availability during a busy week. The more precisely we identify these conditions, the clearer it becomes how far our experience supports our expectation.
The same applies to pain: a person can experience it while being mistaken about its cause. For the question of the experience, that person’s own access matters; the question of a bodily cause calls for further investigation.
The project therefore provisionally distinguishes subjective, intersubjective, and functional-empirical orders. This is not intended as a division into three separate worlds. In our example, our experienced unease belongs to the subjective order, and the agreement as jointly understood to the intersubjective order. When we test the expectation of a quick reply against observable behavior, the functional-empirical order comes into play. These different approaches concern the same situation and help us find reasons appropriate to the question at hand.
How well a claim holds up under the tests appropriate to it is what the project calls validity. How far that support extends is a second question. Our experience so far may support an expectation for particular agreements while leaving other situations open. Likewise, clarifying a commitment we made to one another can establish an obligation without explaining the cause of a missing reply.
The project calls the mismatch between our expectation and the actual course of events friction. It becomes a starting point for inquiry: has the expectation been overextended, the agreement misunderstood, or the situation changed? An answer may lead us to correct a single assumption. It may also show that our general picture has tied different abilities too closely together.
We also need to examine whether we allow contrary indications to count at all. If we find an excuse for every missing reply before asking what happened, we may be protecting our assessment from any correction. If, on the other hand, we treat a single deviation as proof of general unreliability, we give it more weight than it can bear. An appropriate inquiry keeps the question open until the reasons allow us to identify more precisely what the deviation affects.
With a physical model, the same question of what a deviation affects arises in a different form. If a measurement differs from a prediction, we need to investigate whether the measurement procedure, the assumed conditions, or a central model assumption is affected. This determines what needs to be corrected and how broadly the finding applies.
5. When we have to decide before everything is clear
Meanwhile, the event is approaching. We need to establish whether our colleague can unlock the room the next morning. She has done so reliably in similar situations. This time, however, we have not yet received confirmation, and by tomorrow we may still not know exactly why she has not replied.
We now face two tasks. We want to assess her availability appropriately, but we also have to prepare the event. A question that remains provisionally open cannot be resolved by leaving the decision open indefinitely as well. Even doing nothing can have consequences.
The expectation that she will get back to us in time may still be plausible in light of our experience. If the question concerns a suggestion for arranging the room, we could wait and decide for ourselves if necessary. If access to the room depends on her reply, however, we may need an additional safeguard. The same assessment can underlie both decisions; the difference lies in the consequences of being wrong and the alternatives available to us.
We could ask through another channel, involve a second person with a key, or agree on a deadline for confirmation. This can limit our practical dependence while the inquiry into our expectation continues. Making alternative arrangements as a precaution does not amount to a final judgment about our colleague.
The part of the project concerned with the reasoned assessment and use of models is called Epistemics. It keeps two questions separate: what justifies an assessment, and what justifies using it? In making a decision, we need to consider possible consequences and available alternatives as well as past experience. Whether we can correct a step later also becomes important.
If further comparable situations showed that we could not expect a quick reply, we would have to give up that expectation. Agreements that had already been confirmed might still be dependable. For future events, we would then have to wait for a firm commitment or make other arrangements. The fact that no one else is currently available does not make the uncertain expectation more reliable.
A revision changes our model or its claim. In our colleague’s case, this may mean distinguishing availability from keeping agreements already confirmed, and reshaping our expectations accordingly. How far the change should go depends on what the inquiry has actually shown. A small finding can affect a central assumption; a striking deviation can be confined to a particular situation. The extent and direction of the revision therefore need justification, and the revised view must prove itself again.
A revision need not completely displace the previous model. We often retain and use several models alongside one another. To move around the room, a rough sense of proximity and distance is usually enough; to arrange it precisely, measurements may become crucial. Which use is appropriate also depends on time, costs, and the possible consequences of error. A model does not become more accurate because it is inexpensive, but it may suffice for a limited task. Several approaches can be combined as long as their achievements, limits, and possible contradictions remain visible.
An institution, too, has to connect the assessment of a model with the justification for its use. A predictive model may produce good predictions under particular conditions. If it is to help determine who receives support, further questions arise: what consequences do errors have for those affected, and can a decision be corrected? Tested predictive performance and the justification for this use require different reasons. As with access to the event room, the question is what role a justified but limited description may take on.
6. Staying open without losing our footing
Let us return to our colleague. Clarifying the situation can change how we work together. A conversation may reveal that she considered our question less urgent or understood a responsibility differently. It may then be enough to agree more precisely on tasks and reply deadlines for the next event. But a recurring problem may also emerge that cannot be resolved with an extra message alone. How we deal with it depends on both people’s experiences and the conditions under which they work together.
Finding our bearings together does not require complete agreement. We can dispute the meaning of an earlier agreement and still work to clarify it, as long as it remains clear what we are referring to and how the other view can be taken into account. An objection must be able to change something: our recollection, the interpretation of a commitment, or the terms of the next allocation of tasks.
If every other perspective is dismissed as unreasonable from the outset, the conversation loses this possibility. A settled view may then persist precisely because nothing is allowed to count against it. Trust, by contrast, becomes well-founded when we can expect questions and corrections without every shared basis falling apart.
This capacity also matters in larger social orders. Rules and procedures ease our burden because we do not have to negotiate every question anew. But their reliability also depends on whether relevant objections can be taken into account and lead to a correction whose grounds can be understood. The mere fact that an order persists for a long time tells us little about how well it can deal with such objections.
Finally, the picture we have of ourselves also deserves a closer look. The missing reply may have confirmed a familiar thought: “In the end, I have to take care of everything myself.” Such a self-description may rest on real experiences. At the same time, it can influence which tasks we delegate, which support we notice, and which difficulties stand out most strongly. Here, too, it is worth distinguishing the reasons for it from how far they extend.
In this familiar situation, we have seen how ordering experience allows us to make connections and act, and how feedback can change that order. The same questions apply to scientific models and shared procedures. They have a long history: Dewey investigated how an unclear situation gives rise to a directed inquiry. Quine made clear that an experience contradicting our expectations does not, by itself, determine which assumption must change. The project takes up such insights and asks how they relate in the ongoing management of models: which claim is under strain? How broadly does a finding apply? And what follows for continued use?
Epistemics develops these questions into usable distinctions and rules for assessment. Its intended contribution is to connect the assessment of a model’s validity, the justification for its continued use, and its revision. The current core paper elaborates the approach; whether it offers additional help beyond existing procedures remains a question for investigation in its own right.
For the room and the shared event, we usually need less than a complete clarification of every assumption. A confirmed agreement, a justified expectation, and a way to receive feedback may suffice. What we leave open need not prevent us from acting. We need to look more closely when the situation changes or our decisions have consequences we have so far overlooked.
We can thus rely on an understanding without treating it as final. It gives us enough stability to take a next step while remaining connected to experience, other perspectives, and its own consequences. Where this connection is preserved, something that unsettles us can give rise to a more precise question—and that question to a revised orientation that must itself stand up to further experience.