Since Anscombe agency has commonly been thought to involve acting “intentionally,” where intending is often specified as having a “reason for acting” (Anscombe, 1957, p. 9). Furthermore, according to standard “event-causal” theories of agency following Davidson (1963), an act’s being intentional, i.e., its having been done for a reason, means that there is a certain kind of causal relationship between the agent’s mental states (e.g., desires and beliefs) and actions.Footnote 11 The right sort of causal relationship for agency is one where the agent’s intentions are causally connected to its actions, which are meant to bring about a particular effect.
If something like this account of agency is on the right track, then several serious challenges to the possibility of non-successively experiencing agents become apparent. Specifically, I will argue intractable setbacks for NSE agents arise from a faulty causal relationship between the purported agent (the NSE) and the events the agent is meant to cause. Beyond this, NSEs also lack the epistemic status we typically associate with agents. Furthermore, because of their stipulated non-successive mode of experience, NSEs lack the requisite mental states, like intentions, necessary to ground agency. These three problems together suggest NSEs would be incapable of agency given the nature of the world we live in. If non-successive experience is indeed impossible, then we should conclude the Succession Requirement must hold for agents in our world.
Causal Distance and Direction
NSEs face two major obstacles to agency stemming from causality: the problem of causal distance and the problem of causal direction. I will first address the problem of causal distance. Consider a hypothetical non-successively experiencing entity named Abbott.Footnote 12 Abbott, like us, is temporally extended. Only the part of Abbott temporally adjacent to an event such that a causal connection can be made between the organism and the event is capable of directly experiencing it.Footnote 13 The various temporal parts of Abbott thus could enjoy experiences of each time at each time, but there is no way of integrating the experiences had at each time as a diachronic unity spanning Abbott’s life. This is because a causal connection between the parts of Abbott that are temporally distant from the experienced event and the experiencing, temporally proximate parts of Abbott cannot be made such that a unified consciousness over both temporal parts might obtain. This kind of connection would require that events exhibit unmediated and temporally bidirectional causal efficacy, regardless of the temporal distance between the events. However, such “spooky action at a temporal distance” is generally seen as physically impossible.Footnote 14 Short of adding a further dimension to bridge the gap, and thus allowing for a kind of succession in this new dimension, it is difficult to see how the requisite causal circumstances for diachronically unified consciousness over Abbott’s lifetime would be possible in the world as we know it.
To see why Abbott would only have the relevant experiences of a time at the time that the experienced event takes place, it is helpful to consider the eternalist metaphysical picture of time more closely. An NSE is stipulated to exist across and to experience multiple times together, but this would not be possible if only the present exists (as presentism maintains). The eternalist position that we live in a block universe does allow for NSEs, however, as they can then spread out over and experience different times in a fully real sense.
In a four-dimensional block universe, time is not like a landscape to be traversed (Williams, 1951, p. 470). It takes time to perceive, let alone traverse, a landscape. A traversal of time that skips directly over certain sections necessitates another dimension to allow for the process of traversal. Sticking to a four-dimensional metaphysical picture, the experiencing being, which cannot avail itself of illicit “temporal skipping,” needs to be at a particular time to experience events at that particular time directly (i.e., not through memories, imagination, or simulation). However, most of the temporal parts of an NSE are too causally distal to the events the NSE is meant to experience. These parts cannot therefore comprise a diachronically unified subject with one integrated experience.
Besides temporal proximity to causes, we should also expect an agent to be embedded in causal chains in the right way. If an agent were not so embedded and the subject were not affected by worldly causes, its perceptions would be of no use to guide adaptive action, as they would not carry information about the environment. The subject should therefore be embedded such that the effects of the environment arrive successively and then the subject can respond in light of those effects. There is thus a practical impetus to experience things successively: appropriate, informed action is enabled by the sequence of information flowing in from the environment. If part of the function of subjectivity is to integrate disparate information about the world and our bodies to adaptively guide behavior, then experiencing things in sequence is the temporal mode best suited for that function.
Causal direction poses another problem for NSEs. Consider a human translocated to a world where the temporal direction is reversed. The subject’s perceptual apparatus would not be able to pick up signals from the world. The direction of causation would make subjects from our world “blind” to the reversed world and vice-versa. This is a problem for contact between subjects or worlds with differing temporal directions,Footnote 15 but, for similar reasons, it is also problematic for contact between subjects or worlds with a temporal direction like ours and subjects or worlds without a temporal direction at all. If NSEs are understood as not obeying the normal causal directionality of our world, then they would be incapable of meaningfully acting in it. Although this problem fades if we accept that, physically speaking, NSEs follow the normal causal direction, new problems then arise.
Our familiar causal direction turns out to be rooted in the physical nature of our universe, which exhibits causal asymmetry. Craig Callender, following David Lewis (1981), has framed causal asymmetry in terms of counterfactual dependence: “Future outcomes depend upon actions now whereas past outcomes do not,” (Callender, 2017, p. 259). This causal asymmetry is not just an anthropocentric illusion. As Jennan Ismael has argued, when we assess effects of interventions in the physical world from a macroscopic perspective, an emergent pattern of temporal asymmetry arises which is not present in fundamental laws of physics, but which ensures that the “direction of determination” is from past to future and not vice versa (Ismael, 2016, p. 134). As Ismael puts it, “local macroscopic interventions of the kind that correspond to visible human actions affect the macroscopic future but leave the past visibly unaffected” (Ismael, 2016, p. 132). We can then “exploit” these real temporal asymmetries in the activity of the physical components of the universe (which we can call “causal direction”) to act appropriately for our own ends (Ismael, 2016, p. 129). The kind of regularities that we exploit are not mere correlations but related to, as Ismael says, “asymmetries in the way that information propagates from past to future,” which are ultimately grounded in the thermodynamic gradient, i.e., the gradual increase in entropy over time in the physical universe (Ismael, 2016, p. 135; p. 142).
For an NSE, subjective time is stipulated not to have a particular direction, because all times are experienced at once, as opposed to successively. Experienced time would therefore not “flow” forwards or backwards. However, causation, understood as above, remains a feature of the world they and we are supposed to share. The perceptual peculiarities of the hypothetical NSE would still be subject to the direction of causation, underwritten by macroscopic temporal asymmetries resulting from the thermodynamic gradient. While for us, events in the world apparently cause our perceptions of events, in an atemporal world or for a purportedly atemporal being, perceptual experience cannot be so caused.Footnote 16 The subjective experience of the NSE is explicitly atemporal in that it is detached from the ordinary direction of determination, but, like us, their physical existence inevitably succumbs to the familiar causal flow, precluding their stipulated atemporality.
If we think that experiencing a lifetime “all-at-once” means experiencing a range of times at once, then NSEs are confronted by the causal problems outlined above. However, if we think the “all-at-once” experience of an NSE refers instead to a single “static” experiential state that is itself extended over time, then one might think some causal worries can be avoided. Such a state would remain entirely changeless through time, while maintaining causal efficacy at any given time and being the effect of worldly causes at any given time. The latter kind of entity, call it a diachronic-state NSE, is not what our literary examples had in mind, but still represents an important possibility to defuse. Furthermore, one might think we can avoid causal problems by taking the radical step of shrinking our NSE down to the mysterious world of “spooky” quantum interactions, disregarding that agency may well be an emergent macroscopic phenomenon. However, neither of these moves can avoid the further problems with which any kind of NSE must contend, to say nothing of their prima facie implausibility.
In our lives we are confronted by an apparently unavoidable epistemic asymmetry between past and future, in that we have knowledge of the past, but the future is opaque (Ismael, 2016, Ch. 6; Callender, 2017, p. 260). NSEs, however, are stipulated to exhibit epistemic symmetry. I will argue that one result of NSEs’ unusual epistemic condition is that they cannot be said to make choices. If being able to make choices is a necessary capacity for agency, then NSEs are not agents on this count. I will also argue that enjoying such epistemic symmetry is impossible in our world anyway because epistemic asymmetry, much like causal asymmetry, is grounded in the physical structure of the universe. Far from a shortcoming, it is precisely this epistemic asymmetry that enables us to act in ways associated with agency. These considerations lead to the counterintuitive conclusion that NSEs, including the medieval Christian conception of God, would actually be limited in what they are able to do in fairly significant and startling ways. The ways in which NSEs are limited include being unable to deliberate, predict, plan, attempt, and choose, among other abilities that rely on the epistemic asymmetry we take for granted.
In the context of agency, we can say a choice is the result of a decision process selecting between multiple events to make one or more of them come about (Ismael, 2016, p. 152). Making choices is an ability we like to believe we exercise routinely (though of course not all of our actions are choices, e.g., reflexes). An illustrative example is investing. Investment choices are typically grounded in inductive inferences, predicting future performance in light of what is known about the past. Now let’s say that Abbott, as an enterprising heptapod, opens an investment firm. Needless to say, this would be the biggest money-maker on Wall Street. Abbott is, after all, privy to the future to the same extent we are to the past, as he enjoys epistemic symmetry with respect to time. In fact, Abbott’s situation is even better—whereas we rely on corruptible records and fallible memories for information about the past, Abbott can immediately perceive every historical and future price at any given time. This is clearly a recipe for financial success.
However, Abbott is unable to make investment choices because he does not make a decision in order to bring about a later event. As far as Abbott is concerned, the investment is already, at any given time, perceived as invested. Or, at least, that’s the nearest description we can give for Abbott’s fundamentally tenseless experience. There is not even a sense for Abbott that he will (or did) invest in a particular asset as, from his perspective, it would always be as if he is in the moment of buying the asset, forever catching himself in the midst of high finance. We can see that, because of this untensed mode of experience, for Abbott no process of deliberation takes place, no prediction, no decision making, no attempt to bring about an event, and no selection process. These actions all presuppose the utilization of past information in service of enacting future outcomes in an unrealized future. Actions of this type are not only rendered irrelevant by epistemic symmetry but would be impossible except as charade.
This is not merely a problem of omniscience. Knowing how things are going to play out does not necessarily preclude choice. For instance, I might have insider information such that I know tomorrow exactly what will happen, and I may know that I will capitalize on this information by purchasing a certain asset, but nonetheless I will still choose to buy it.
Abbott is in a different predicament. Abbot is not just omniscient but temporally omniceptive (to coin a term), meaning he perceives every time at once. Experiencing all times immediately is a stronger epistemic position than just knowing what happens, because experiential knowledge outstrips ordinary propositional knowledge. For Abbott it would be as if, at any point in time, he sees his portfolio including everything his investment company has ever bought, is buying, or will buy. With the entire temporal extension of the portfolio immediately transparent and immutable, there are no choices available for Abbott to make. If making choices is essential for agency, we can see then that NSEs fail to satisfy this condition. So, while Abbott’s foray into finance might be exceedingly lucrative, it would not be empowering.
An NSE also does not try to do things. Trying requires a desire for an eventuality to come about in addition to the possibility it does not. However, from Abbott’s perspective, trying would be nonsensical. Abbot might for example join an expedition to the South Pole, but he will not try to arrive there because, from his perspective, he is always already there (or not—he might expire en route, in which case, for him, he is already dying in Antarctica from the time he comes into existence).
Likewise, NSEs have no need to deliberate or plan for the future, as they are supposed to experience what we call the future as if it were what we call the present, making planning redundant (although Abbott could of course make a show of assessing an earnings report, going through the motions of an investment decision and fooling his clients). Indeed, NSEs cannot deliberate or plan because it is impossible for them to perceive an open future. From an NSE’s perspective, there is simply nothing to plan for.Footnote 17
It is likely that NSEs lack many more capacities often associated with agency that rely on epistemic asymmetry as well. One of these is the ability to “temporally decenter,” or take a temporal perspective that is different from one’s own (McCormack and Hoerl, 1999, p. 174). Craig Callender, as well as L. A. Paul, maintain this temporal decentering ability is crucial to the emergence of the self, understood here as a “narrative center of gravity” (following Dennett, 1993, p. 418), and crucial to the emergence of an agential perspective (Callender, 2017, pp. 255–259; Paul, 2017, pp. 263–264). Reflection on an NSE’s circumstances reveals they would be deficient in this regard. For a being that experiences its entire life at once, there can be no question of taking different temporal perspectives because that being takes all of its possible temporal perspectives at any given time. There is no picking and choosing moments of its life to experience. The heptapod cannot revisit its “past” or imagine its “future.” Being able to do these things requires successive experiences.
A related issue connected to the NSE’s “closed future” is that, as Callender points out, “All of our evidence is confirmatory of the idea that our decision is the causal trigger that leads to, or brings about, the event. Anything prior to that decision can be trumped by the decision itself. […] Part of what it is to be an agent is to have this sense of freedom, a sense that other future options are in some sense live,” (Callender, 2017, p. 260). This same feeling does not apply to our past, about which we cannot change our minds and expect a different outcome. However, for an NSE there is no special epistemic status for the future. It is experienced like any other time. It is therefore plausible that, for an NSE, there would not be a sense of agency either.Footnote 18
A small wrinkle appears in these epistemic considerations, however, if a finite NSE’s action takes place beyond the scope of its experience, e.g., after death. Infinite NSEs, like the medieval Christian God, would be omniceptive over eternity, but finite NSEs, like heptapods, could still be ignorant of those things which take place after their life ends. Thus, regarding circumstances beyond the scope of their total experience, an NSE runs into the same epistemic asymmetry we do and consequently may appear more like an agent, albeit from beyond the grave.
Even so, there are still reasons for thinking any NSE’s epistemic setup just wouldn’t work in our universe. An NSE faces no pressure to gather information to guide its action and yet it is this task our minds appear to have evolved specifically to carry out. The reasons we face an epistemic asymmetry are also not an accidental feature of our human minds. Once again, there are physical facts constraining us. As Ismael puts it, “what explains our greater knowledge of the past than the future is that along [the thermodynamic] gradient, inferences from the present, surveyable macroscopic state of the world to its past […] are much more powerful inferences than inferences from present to future” (Ismael, 2016, p. 143). Ismael goes on, “the thermodynamic gradient makes it possible to create records in the environment that will carry information to our future selves” (Ismael, 2016, p. 145). What the gradient does not do is make it possible to carry information from our future selves to our past selves, which is what an NSE needs to do to exist as a diachronically unified entity. To us, the future is not only epistemically inaccessible but seems as if it has not yet occurred and will come to be shaped by us. The propagation and accumulation of information apparently works in one direction, such that even if the future already exists in some sense, it might as well not to any being in our universe. This is a potentially insurmountable obstacle to the actual existence of an NSE. An upshot of this is that our own successive mode of experience should be seen not as limitation, but on the contrary, an enabler of a kind of existence that is uniquely suited to thrive in the world we live in.
As mentioned previously, on standard accounts of agency, agents need to have particular kinds of mental states. Specifically, agents need to be able to form intentions and act on them, and these intentions should be causally related to the eventual actions. I will now argue non-succession precludes the formation of intentions, thus ruling out agency as traditionally understood. NSEs should be seen as incapable of forming intentions for two main reasons. One is a fairly straightforward worry related to the problem of epistemic asymmetry. The worry is that, while we intend future events to come about as a result of our actions, the future is not available to an NSE as a domain of uncertainty. It then does not make sense for an NSE to intend that future events come about, as these events have no special status and are experienced directly. Just as we do not form intentions concerning the past, NSEs would not form intentions about the future.
A second, more troublesome worry is rooted in the ontology of mental states like intentions, itself intimately related to the ontology of experiences. Experiences are not static, unchanging entities, but evolving processes defined, at least in part, by their dynamism. To make this thought more ontologically explicit, we can follow O’Shaughnessy (2000, Part I.2) and Soteriou (2016, Ch. 6), who claim experience is “not merely a continuous existent across time, [but] an activity and therefore also a process, and thus occurrently renewed in each instant in which it continues to exist,” (O’Shaughnessy, 2000, p. 42).Footnote 19 Occurrent processes are defined by being successively present (Soteriou, 2013, p. 139). Unfolding processes, as opposed to states, also allow experienced events to have duration, i.e., to take time. If we agree that experience should be defined in this way, then an NSE is incapable of experience as such. This is because, for a non-successively experiencing being, every event of its life is supposed to be experienced all at once at each moment. There can then be no individuatable experiences at different times, separately caused by different events or that cause different events at different times. Instead, there would be just one state.
While the above calls into question the very possibility of non-successive experience, one still might think a non-experiencing entity could be an agent. It is perhaps still possible with this understanding that NSEs have intentions, provided intentions are not necessarily processive. If an intention is just a set of dispositional facts, for instance, this could be the case. Abbott could still exhibit an innate set of non-qualitatively felt dispositions that occupy a causal role(s). However, such dispositional properties do not seem distinctive of those entities we consider to be agents, nor relevant to their status as such. Inanimate objects like toasters also have dispositions to behave in certain ways but seeing toasters as intentional agents does violence to our ordinary concept of agency (cf. Dennett, 1987).
I would argue instead that mental states like intentions must be occurrent processes. One reason for thinking this is the case is that intentions have a distinctive cognitive phenomenology—it feels like something to intend to do something, and it feels like we are the ones doing it.Footnote 20 There is an urge towards action preceding and ostensibly precipitating the action. For genuine agents, it is not just that we act because we are disposed to do so, even if dispositions play a role. Mental states like intentions are, at least in part, experienced, and as such, must be processive for the same reasons as experiences. Without this experiential component, the NSE lacks a critical element involved in our understanding of intentions, and thus also fails the intentional criterion for agents.
Beyond the inappropriate cognitive phenomenology, the changeless intentions of an NSE would be unable to play the role of cause to all the actions of the subject. In order to do so the state would have to exhibit differing causal powers and susceptibilities at different times, which it can’t do without changing in some way. However, for the NSE, there can be no change in intention, as a change in intention necessitates an experiential succession.Footnote 21
Now, it may appear this problem is avoided if a mental state can change without the NSE’s experience changing. So long as we accept there are subpersonal mental states such a contention seems eminently plausible. However, the kinds of mental states that are most relevant to agency are not subpersonal ones. It is personal-level mental states that must have the appropriate causal role, otherwise we are hard pressed to distinguish automatic reactions from the actions of an agentive subject. As a result of these considerations, we should think NSEs are incapable of forming dynamic intentions and thus also incapable of the kind of intentionality that agency requires.
One might think there may yet be ways of salvaging agency for NSEs. For instance, NSEs might have sui generis non-processive intention-like states that occupy the functional role of an intention without phenomenal character. These non-processive, non-experiential states might be causally related to actions and their effects. Limited to a changeless state extended over time, and correspondingly limited in causal power, without cognitive phenomenology, experience, or the various abilities predicated on epistemic asymmetry, such an entity would stray radically from our ordinary understanding of agency.
NSEs and Non-standard Theories of Agency
We might still wonder how NSEs would fare on a non-causal approach to agency (e.g. Melden, 1961; O’Connor, 2000). On such accounts, the agent need not stand in any particular causal relation with their actions. Understanding agency in this way doesn’t let NSEs off the hook for several reasons. One is that the problems of causal asymmetry are not only problems for agency, but also problems for the NSE’s existence as a diachronically unified subject. There just cannot be direct causation between temporally distal events or non-directed causation on the macroscopic scale, yet these conditions must obtain in order to furnish the NSE with the experience it is stipulated to have. However, even ignoring these causal issues, an NSE would still run into epistemic asymmetry problems and the problem of non-intention.
Suppose however we agree with Anscombe (1957, p. 9) that intentional actions are those which are done for a reason, but reject the standard contention that agency also requires a special causal connection between intention and events. Perhaps then, it might seem, NSEs can at least avoid the problem of non-intention. We might think a non-successively experiencing being may still have reasons to act in certain ways, even though such reasons may have to be as “ever-present” to the NSE as the NSE’s experience.
I do not think such a move would be successful. Appealing to reasons requires the subject enjoy certain mental states, including belief states. However, as discussed earlier, these mental states have a distinctive cognitive phenomenology that suggests they must be processive in character and so not the kind of thing that can be had at a single moment. Instead, it is essential that beliefs are experienced successively. Perhaps more importantly, though, is that for an agent to act according to reasons, the reasons should precede the action. Unfortunately, non-succession makes this impossible by definition.
Losing Your Self in the Moment
So far we have been considering experiential succession as a precondition for agency, but it is likely experienced succession is also a condition for selfhood. While I think this idea is somewhat intuitive, it is not universally shared. J. David Velleman, for instance, drawing on Buddhist thinking and Derek Parfit’s “neo-Lockean” account of persons (Parfit, 1981, p. 281), maintains “the existence of an enduring self is an illusion” (Velleman, 2015, p. 175). For Velleman, however, the emphasis is on enduring. An enduring self persists through time in such a way that it is whole at any given time. Though this might be how we ordinarily conceive of ourselves, Velleman argues selves are in fact not like this at all. Rather, selves should be seen as perduring objects, in that they persist by the succession over time of individual temporal parts.
Recognizing ourselves as perduring entities, Velleman hypothesizes, could induce a radical and potentially liberating perspective shift. As he puts it: “Suppose that I could learn to experience my successive moments of consciousness—now and now and now—as successive notes in a performance with no enduring listener, no self-identical subject for whom these moments would be now and then and then again” (Velleman, 2015, p. 187). Your present self, in this scenario, “would think of itself, and each of the subjects with whom it communicates by memory and anticipation, as seeing its own present moment, with none of them seeing a succession of moments as present,” (ibid.). This present self, and indeed each momentary self on such a conception, should by now look very familiar. What we have here is a collection of non-successive experiencers together comprising what we mistakenly think is an enduring self.
Velleman goes on to claim that for one’s present self, “time would no longer seem to pass, because [one’s] experience would no longer include a subject of its passage—just successive momentary subjects, each timelessly entrenched in its own temporal perspective,” (ibid.). As tentative evidence for this, Velleman cites research on “flow-states” during which we tend to “lose awareness of time’s passing” and lose self-awareness to boot (Csikszentmihalyi, 1990). Velleman is concerned with the extent to which we can shift our perspective towards the perduring self and whether this could console us and alleviate suffering. While Velleman ultimately admits that, as far as we ordinary humans are concerned, “we can’t stop the self from seeming to endure,” (Velleman, 2015, p. 192), I would argue the thought experiment itself is incoherent, insofar as it presents NSEs not only as possible but as our fundamental constituents.
We have already seen that NSEs cannot be agents. We would therefore be remiss in applying agential language to the time slices of Velleman’s hypothetical scenario. The notion of self does not apply here either. If we identify the agentive self with a “locus of control,” (Ismael, 2016, p. 101), then without successive experience such a self does not exist, given the arguments of previous sections. Other conceptions of the self also fail to obtain without successive experience. Consider, for example, that there are a number of essentially processive actions, such as “seeing,” “thinking,” and “communicating” which can’t be said to be things that a momentary subject is ever capable of doing. There are predication errors in describing a momentary self as interacting with its environment, itself, and other subjects. The reason for the errors in predication is that the appropriate grammatical subject to which the predicates would be applied has dropped out of existence by failing to extend over time. We need that subject to exist wholly at multiple times, not only at one time, in order to predicate of that self the processive actions we typically want a self to be able to perform.
A more accurate description of the consequences of fully realizing the perduring self would be an absence of an experiencing self. Non-successive experience is not just nomologically impossible for agents, it also entails non-selfhood. This does not mean a state of non-succession isn’t possible; it may be, but only if we eliminate both selves and our agency. This sort of total “ego-death” is reported by users of hallucinogens (see, e.g., Nour et al., 2016) and is arguably the goal of some forms of meditation. Many Buddhist traditions,Footnote 22 for example, do not seek a scenario like Velleman describes, with its multiplicity of selves, but instead seek “no-self” (Sanskrit: anātman).Footnote 23 If the truth is that we are perduring rather than enduring entities, fully realizing this truth could then be more world-shattering than Velleman’s thought experiment suggests.
There are potential ethical consequences here as well. If SR is true, as I have claimed, of both agency and selfhood, then it is likely NSEs would not be appropriate targets of moral judgment. Because non-agents do not have a claim to moral responsibility, it would make as much sense to call an NSE’s actions “good” or “bad” as it would to describe a toaster this way. Taking the thought to its most provocative conclusion, if we consider our most famous alleged NSE—the medieval Christian God—it seems we have to say claims like “God is good” involve errors in predication as well.
It should be plain that I am not claiming selves or agents necessarily exist in any strong sense. What I am claiming is that, for such things to exist in an ontologically strong sense or as some sort of illusion,Footnote 24 it is paramount that we as subjects enjoy a successively structured experience. There are, of course, very many different conceptions of “self,” and exploring the role that the temporal structure of our experience plays for each of these exceeds the scope of the present paper. However, I hope this brief excursion has highlighted the potential extensibility of SR to domains beyond agency.