In this post, I introduce a distinction between first-order future possibilities and second-order future possibilities and discuss why it is important.
There are two ways of mapping the space of future possibilities:
(1) We can take some view on the world (for example, a theory about some domain or a personal idea of the workings of some domain, etc.) and construct scenarios on the basis of that worldview. In this method, we first explicate a set of principles that govern some domain. We then make assumptions about specific states of affairs that might occur and then track the consequences of these states of affairs by using the principles. (See detailed discussion here.) We then arrive at a set of scenarios that describe a set of possible futures.
This method provides what I call a structurally unified set of scenarios. All the scenarios that are constructed on the basis of one view on the world share the principles and these principles set the limits for what is possible. The space of possibilities is therefore fixed by structural consideration provided by the view. This does not mean, however, that there are no disagreeing scenarios in the set of scenarios provided by one worldview. Different claims and interpretations about states of affairs may change the scenarios that are produced.
Take an example. In philosophy of science, scientific realism[1] says that the mathematical contents of successful theories are preserved through theory change. This view generates a set of scenarios where there will not be radical changes in successful scientific theories. The set of scenarios is structurally unified in that it does not allow certain scenarios (i.e., scenarios of radical change is successful science). However, there might be disagreeing scenarios about the future of some particular scientific theory, depending on whether is considered successful enough. How we judge a state of affairs (i.e., whether a theory is successful enough) affects how we see its future possibilities.
The first method provides first-order level future possibilities: They are possibilities that follow when we remain coherent with respect to the view we accept. They could also be called internal possibilities: they are possibilities that follow internally from a view on the world, i.e., they follow from the internal principles. The first-order level tells us what is possible, if our views are correct. First-order level possibilities are epistemically tractable by assessing the view on the world behind them.
(2) We can produce a richer set of scenarios by using many views on the world at the same time. Each view generates its own set of possibilities and therefore combining them usually leads to a larger set of scenarios. However, the scenarios that two views generate may be overlapping. If there is no difference in the scenarios, this is an interesting result in itself: two seemingly different views agree on what can happen in the future.
For example, we might have the view that radical changes occur even in successful science. This view would generate a set of scenarios among which there are scenarios where successful science goes through radical changes. Used side-by-side with scientific realism, the two views would generate a rich set of scenarios.
The second method provides second-order level future possibilities: They are possibilities that follow when we distance ourselves from a particular worldview. They could also be called external possibilities: they are possibilities that cannot be generated with only one view on the world, i.e., their construction requires that we step out from any particular view. The second-order level is, indeed, best suited to challenge our assumptions and to see what is possible, if we are wrong. The epistemic assessment of second-order level possibilities is a non sequitur because the mutually incompatible views that generate them cannot all be justified (by a given set of evidence). We cannot justifiably believe that all the scenarios are plausible.
This distinction has some interesting consequences:
First, two persons could agree on particular future possibilities without being aware of the fundamental structural differences in their maps of future possibilities. If one person believes in principles P and states of affairs A, and another person believes in principles P* and states of affairs A*, they might end up in the same scenario. For example, a realist might believe that successful science does not go through radical changes (P) and that a theory is not successful enough (A). The realist would then believe that the theory can go through radical change. On the other hand, “radicalist” might believe that all science goes through radical changes (P*) and that the theory is successful (A*). The radicalist would then also believe that the theory can go through radical change. What the radicalist claims is not possible in the first-order level of the realist, given A*. For the realist, what the radicalist claims is possible only on the second-order level. These two levels should not be conflated due to the seeming agreement. One main reason for this is an epistemic one: The realist could recognize the radicalist’s possibility but could not epistemically endorse in as long as she thinks that realism is well justified.
Secondly, and as a corollary of the previous point, a mere “brute” set of scenarios is not very informative about future possibilities if we do not know the levels of different scenarios. The brute set might make us able to recognize a wide range of possibilities without any idea of how to assess the plausibility of a particular scenario. Prima facie, as long as the generation of the scenarios is not transparent, we should assume that the set of scenarios describes second-order level possibilities. Only if we can group scenarios together on the basis of the views that are used to generate them, we can identify first-order level possibilities (relative to a view). Only by identifying first-order level possibilities, we can assess the epistemic ground of the possibility. This is a pressing problem for every approach that constructs scenarios on the basis of what some people think about the future: Different people may have different views and they may have illusionary agreement on what is possible (i.e., confusion about first-order and second-order level possibilities). The epistemic assessment of scenarios or even the structure of the space of possibilities cannot get off the ground before the views where different scenarios stem from are systematically explicated. In practical approaches that focus on what some people think about futures, systematic explication is often missing. We are stuck with a brute set of scenarios that must be interpreted as an epistemically inaccessible set of second-order level possibilities.
Thirdly, while the first-order level is the only level where the possibilities can be, strictly speaking, epistemically assessed, the second-order level is highly useful. It can challenge first-order level judgments. It can also serve us in situations where our confidence in our views is low. In such cases, it allows us to answer the question “what could happen, if we are wrong?”. However, a question like this cannot have a justified answer because justification flows from what we believe and assumes that we are not wrong.
Finally, the distinction reveals two fundamentally different ways to reason about possibilities. On the one hand, one may follow one’s convictions and see what possibilities follow. A responsible agent would also assess the merits of the convictions. On the other hand, one may take distance to one’s convictions and figure out what possibilities – alternative possibilities – would follow from other convictions. However, what is an alternative possibility to one person is a mainstream possibility to another person. There are no “alternatives from nowhere”.
In conclusion, justifying and reasoning about future possibilities requires that we distinguish between first-order level and second-order level possibilities.
[1] I refer to structural realism but omit the “structural” to avoid confusion with the other use of structural in the post.