How arrogant is it to call a special issue ‘new trends’ in anything? How could you possibly hope to pinpoint, among the recent changes of a field, those that will prove to be momentous? We couldn’t and we haven’t. But this does not matter, for rational choice theory (RCT) is a peculiar beast. Although standard RCT is often equated with the principle of expected utility maximization, it has always been something of a hotchpotch, a mix of various approaches that maybe share no more than a family resemblance. Accordingly, there’s long been a variety of opinions regarding whether and in which way RCT should be developed, enriched, amended or extended.

This special issue, while not (possibly) being exhaustive about the recent contributions made to RCT, reflects the variety of reactions and worries made to this original version as well as to some of its later versions. Maybe standard RCT is not justified enough—some of its basic tenets may not be considered as rational requirements. In the past, most such worries have concerned basic preference axioms. Here, it concerns time-discounting—the fact that we prefer something now rather than the same thing later (Ahmed).

Maybe RCT is based on too many presuppositions, some of which should be turned into targets of inquiry. Recent applied decision tools do just that, which raises the question of their complementarity with standard approaches (Helgeson).

Maybe, because RCT provides inadequate answers to specific problems, it should be extended, complemented or modified. In particular, the question of how to treat risk into decision framework (Stefánsson, Weirich) and/or of how to decide how to set up a space of outcomes from which decisions are assessed (Mamou).

Maybe RCT should be further extended to cover agents that are not standard individuals, for instance groups of agents. Such approaches have been around for a long time, but the question of how to best understand and formalise it is still relevant today (Gold & Colman, Tummolini & Stirling).

Finally, maybe RCT should be abandoned, given the numerous harsh criticisms that it has faced almost since its inception. On the other hand, the very diversity of versions of RCT may forbid any general dismissals and at best allow for targeted, context-dependent criticisms (Herfeld).

Overall, this set of contributions encompasses the many ways in which RCT remains relevant today—even after all this time, its basic tenets are still discussed; at the same time, it is not disconnected from recent developments.

Let’s be more specific about the contributions to this special issue. Ahmed’s paper targets a possible justification of pure time-discounting, namely Parfit’s idea that we care more about a nearer future self’s well-being because of the more numerous psychological connections we have in common. Ahmed shows that this psychological justification, once made precise under reasonable assumptions, entails irrational kinds of time-discounting. In other words, time-discounting cannot be justified under at least one of its prominent psychological interpretations.

Helgeson discusses so-called ‘deep uncertainty’ approaches, used when probabilities regarding actions or outcomes are unknown, and the way in which it relates to classical decision theory. The latter typically deals with rationality or decision rules, on the basis of predefined sets of states and of acts. By contrast, deep uncertainty approaches concern the very structuring of decision problems, that is, either the partitioning of state spaces or the forming of menus of acts, both of which are typically neglected by classical decision theory.

Stefánsson describes a life-and-death scenario in which classical decision theory provides counterintuitive recommendations and argues that the best way to justify the intuitive move is to have agents ascribe different utilities to changes in risk depending on the initial risk level. Stefánsson then introduces a formal framework which allows for this, and in which propositions about chances may receive utilities just as propositions about outcomes.

The next two papers deal with the way in which decision problems should be structured. Contra one of John Broome’s arguments, Mamou argues that avoiding the intransitivity of preferences by re-individuating the available options does not render the transitivity principle empty. Though always possible in principle, it should depend on what is deemed relevant in the decision situation at hand—which lies outside the scope of decision theory.

Weirich gathers both Stefánsson’s and Mamou’s topics. Like Stefánsson, he holds that rational agents should care about risk, although in a different way for different reasons. If RCT is to explain the preferences of agents, then an option’s risk should feature in this option’s consequences—even if measuring it may be difficult. This allows one to make decision theory compatible with observed choices in problematic situations such as versions of Allais’ paradox.

We then have two papers pertaining to group rationality. Gold & Colman discuss whether it may be rational to choose a payoff-dominant outcome in a game—a long issue when investigating solutions for coordination games or social dilemmas such as the Stag Hunt. While they review numerous possible solutions, their focus is on team reasoning, which does warrant payoff-dominant outcomes. However, it does so by presupposing that agents identify with a group, a process which can hardly be made rational in turn.

Tummolini & Stirling highlight quite a different formal option for rationalizing coordination (i.e. justifying the choice of collectively preferable outcomes). In their framework, agents may influence each other at the level of preference formation and so have conditional preferences—one’s preferences may depend on one’s conjectures regarding the others’. A number of technical results ensue, a striking feature of which is that preferences then behave like probabilities. The nature of the social influence at the preference level then allows one to explain efficient coordination (among other things) without positing any kind of collective agency.

Herfeld’s paper concludes our special issue with a wide-scope review of the various criticisms that have been made to RCT, in order to argue that none of them is fully general. This is because RCT is best understood as a disunified field, the core concepts of which (rationality, utility, preferences, etc.) may receive various definitions. Indeed, it is best understood as a set of research strategies centered on some concept of rationality in general, whether for empirical, explanatory or conceptual concerns. Any criticism is bound to be piecemeal, that is, dependent on a specific variant of RCT and on the context in which it is used.