Does moral responsibility require mental time travel ? Considerations about guidance control

The debate about moral responsibility for one’s actions often revolves around whether the agent had the ability to do otherwise. An alternative account of moral responsibility, however, focuses on the actual sequence that produces the agent’s action and which criteria it must fulfil for the agent to be considered morally responsible for her action. Mental Time Travel allows the agent to simulate a possible future scenario; therefore, it is relevant for the selection of a course of action. I will argue that implicit prospection is a rudimentary form of Mental Time Travel and that the role that implicit prospection, or non-rudimentary forms of Mental Time Travel, plays in the production of intentional actions helps explain guidance control and, hence, moral responsibility.


Introduction
Recent discussions in the cognitive sciences have considered the relevance of Mental Time Travel (MTT) to moral agency (Ger ans and Kennett, 2017).Here I consider discussions on moral agency in an admitte ly modest sense compared to other ap roaches; my concern is whether moral responsibility can be attributed to the agent for at least some of her intentional a ions. 2 Considering that MTT is currently we l accepted in the cognitive sciences (Tulving, 1985;Su dendorf et al., 2009), here I wi l examine whether MTT is a reasona le requirement for moral responsibility for one's a ion.Ger ans and Kennet (2017) think so.I wi l ar ue that implicit pro ection, which is not yet explicit MTT, is not in fact a criterion itself; nevertheless, it is relevant to the criterion for moral responsibility-uidance control-and helps explain it.
I start by making the distinction between implicit pro ection and explicit MTT and by presenting Ger ans and Sander's (2014) ar ument for the possibility of implicit pro ection.In the next section, I wi l show that we l-known accounts of intentional a ion are compati le with the requirement that implicit pro ection, and sometimes MTT, play a relevant role in the production of such a ions.Then, I discuss Fischer's uidance control, its requirements for moral responsibility, and how implicit pro ection helps ground moral responsibility for typical humans when they act.My aim is to show that implicit pro ection contributes to further explain uidance control's requirements.Fina ly, in the last section, I consider a possi le objection.

Implicit prospection and intentional action
It may be useful to briefly describe Ger ans and Kennett's (2017)

account of MTT:
Mental time travel simpliciter consists in the simulation of perceptual and sensory aspects of episodes of future experience, for example by imagining the view from a mountaintop, or the taste of taro.Dynamic evaluation consists in the association of affective responses with those simulations in order to aid decision-making (Gerrans and Kennett, 2017, p. 261).
It has been advanced by simulation theories that MTT is one capacity of the system that ena les human constructive episodic imagination (past, future, couterfactuals) in general, as we l as mind reading, and episodic conterfactual thought (Michaelian, 2016).According to Michaelian (2016), MTT can be directed to the past or to the future; roughly, the distinction is that, when directed to the past-episodic memory-the subject (constructively) simulates an episode from her personal past, while future directed MTT simulates (imagining) a possi le future (often for herself, though this kind of simulating can be nonpersonal).De Brigard (2014) espouses a related view, which considers memory and MTT part of the Episo ic Hypothetical Thinking system.Scha er and A dis's (2007) view emphasizes the constructive character of memory and the similarities between its neural substrates and those of imagining the future.
Episodic memory3 a lege ly assimilates one's experience of oneself into, say, a memory; i.e., one does not just remember one's chil hood home, but one remembers what it was like to be in that place.Autobiographic memory has a strong affective component, the affective valence.Imagining the future is in this sense similar to episodic memory, according to Ger ans and Kennett (2017).Imagining future possibilities presents the agent with alternative courses of action (even consequences of these actions) from which she may choose; MTT is an affectively enriched episodic representation used prospectively (Ger ans and Kennett, 2017, p. 262) that uides action selection as part of action production.
The capacity to project oneself in future a ions is essential to form a ion preferences, a plan.The capacity to expe ience onself in the future is, hence, crucial for settling on an a ion plan.Thus, the phenomenology of MTT is one of its relevant a ects.It accounts for the distinction between plans that have an affective component, say, there is a distinction between thinking about a friend's plan to travel to Patagonia and thinking about my plan.When the agent plans to travel, she imagines herself experiencing her goal, which explains the affective a ect of the plan; she sees herself in the trip.
Phenomenology contributes to explain why MTT such as planning is not motivationa ly neutral, because, roughly, the agent expe iences herself in the future, and when one looks to the future instead of the past, the affective valence towards alternative courses of a ion inclines the agent to act (Ger ans and Kennett, 2017, p. 261); i.e., it is not motivationa ly neutral.
Given that MTT is said to be exclusively human, MTT's phenomenology is the main a ect that distin uishes human MTT from more rudimentay forms of episodic-like memory found in other animals.In the case of implicit pro ection, the phenomenal a ect is no more than feeling the future.Ger ans and Sander ar ue for conceptual space for implicit pro ection.Roughly, they define what they mean by implicit pro ection as fo lows: "there are forms of pro ection that can be uided by implicit simulation of information represented in past and future experience" (2014, p. 701) which are not fu ly under the agent's executive control.They are particularly concerned about whether these implicit representations contribute to uide a ion.The method they use to provide conceptual space is showing that implicit pro ection, as they conceive of it, is consistent with the availa le neurocognitive and behavioral evidence, and the hypothesis they defend is that implicit pro ection is an early processing stage that ena les explicit MTT.Thus, the neural substrates for both these processes would be the same.The advantage of this hypothesis is that it does not postulate that a ditional cognitive mechanisms or processes are necessary for implicit pro ection since it would be an earlier, rudimentary, stage of an already we l-accepted process, explicit MTT.
The ar ument for implicit pro ection relies on an analogy to the early development of cognitive capacities that I wi l not discuss here and on cases of prosopagnosia.The ecific kind of prosopagnosic patients Ger ans and Sander (2014) are concerned with cannot explicitly recognize faces, but they seem to be capa le to react to familiar faces based on, a legedly, the affective valence (emotional significance) of an implicit representation of the face.Prosopagnosia patients, a lege ly, implicitly, but not explicitly, recognize faces, which a lows for affective response to plans involving familiar and unfamiliar people, as we l as friends and enemies.Ger ans and Sander ar ue that these are cases of implicit representations uiding a ion.Moreover, if these patients were to choose dining companions for a future dinner taking into account only their names, Ger ans and Sander (2014) ar ue that they could proba ly do so based solely on the affective response to implicit representations, and this would be a case of implicit pro ection.To defend that this is a case of implicit representation they ar ue that "the model of face recognition which structures research in the area depends on the idea that familiarity of faces is computed in order to produce the kind of behavioural effects we are intere ed in.In other words the relevant information is implicitly represented" (Ger ans and Sander, 2014, p. 706).This is relevant to their point because if the agent can simulate a situation that is not cur ently happening in her environment to uide which future-oriented a ions she wi l perform, then this is enough to e a lish that this is a case of pro ection, and in the case of prosopagnosics, that this is a case of implicit pro ection.
Another evidence is derived from the Iowa Gam ling Task (IGT).Ger ans and Sander (2014, p. 707) claim that implicit representation of the punishment schedule (pattern of gain and loss represented by the cards) in the IGT antecedes anticipatory skin conductance response (SCR) and explicit knowledge about the schedule.Ger ans and Sander accept that SCR depends on implicit representation; after a learning period, the subjects' choice of card reflects the punishment schedule even before they can explicitly describe the schedule.Therefore, they claim that implicit representation is the relevant representation in these cases.SCR is considered anticipatory in the sense that it uides future a ion; therefore, it can be considered as pro ection, and they claim that this is a case of implicit pro ection.
Furthermore, Ger ans and Sander propose that, in case there is sti l any doubt, if subjects can merely imagine themselves drawing a card form one of the decks, then this would definitely be an instance of pro ection (simulation).Proba ly if this were done before the subject had explicit knowledge about which are the advantageous decks, then it could be considered an instance of implicit pro ection.They conclude that "[…] if an organism can generate and use those implicit representations in the absence of the stimulus she is effectively remembering or imagining the represented object" (Ger ans and Sander, 2014, p. 708).
In a dition to ar uing that they have made conceptual space for implicit pro ection, Ger ans and Sander (2014) claim that this kind of pro ection is enough for the agent to feel the future4 and for the cost of alternative courses of action become affectively salient.In such a scenario, processes of implicit representation would determine a rank of (affective) preferences for alternative courses of a ion, provided that there is a sufficient period of indecision (which may be very short).This explains the development of preferences for implicitly represented future a ions and outcomes based on affective responses to these.
If this is cor ect, then there is conceptual space for the notion of implicit pro ection, which may help explain the production of intentional a ions.5

Intentional action
One prominent chara erization of intentional a ions6 considers that they are done for reasons to act-which explain a ions-; these reasons contribute to the acquisition or formation of the intention that plays a role in the production of the intentional a ion (Mele, 1992, p. 109-115).My focus here wi l be on intentions; more precisely, on a ion plans taken as a crucial element of intentions (Mele, 1992;Bratman, 1987).In Mele's terms, intentions are motivation-encompassing attitudes that commit the agent to the a ion (Mele, 2003, p. 28).Moreover, an intention commits the agent, because it is an executi e attitude towards a plan (Mele, 1992;2003, p. 27).I believe a connection can be drawn here to MTT and implicit pro ection, because it can be said that planning is an instance of MTT or, at least, that MTT contributes to the selection of an a ion plan.
I wish to advance the thesis that the plan encompassed by the intention that plays a relevant role in the production of an intentional a ion (Mele, 2003) is an instance of MTT.This would help explain both intentions' future direction and their motivational a ect.The said motivation may be partia ly inherited from a cur ent desire to do A; nonetheless, it may (partia ly) be encompassed by the a ion plan, that is, the pro ection itself.The agent a lege ly feels the future when she plans to do A; i.e., her planning involves experiencing herself in the future she wi l inhabit as a result of the planned a ion course.It can be ar ued that the plan has an affective valence toward the planned future episode in which the agent wi l find herself, if a l goes as planned.In fact, the affective valence of the simulation can affect its phenomenal quality, i.e., how intensely it is felt (De Brigard and Giovane lo, 2012).Experiencing herself in the planned future wi l stress the cost of the agent's course of a ion; in other words, the affective valence of the possi le future wi l be ap arent, which may be affectively positive or negative.
It is reasona le to assume that countless courses of a ion are open possibilities to the agent at each moment (I wi l not get into the discussion about whether they in fact are open or just ap ear to be).If one accepts that MTT contributes to the process of selection of one of these courses of a ion, then MTT is part of the production of most intentional actions.This is proba ly not the case for automatic a ions, and less so for reflexes, given that these are produced by means of less flexi le processes.MTT plays a role when there is at least some minimal indecision about which course of a ion wi l be selected (when there is some flexibility), even if just for a very short moment.
This proposal, nevertheless, could strike one as unrealistic, considering that typical human agents do not seem to engage in explicit planning every time they act intentiona ly.Considering such a scenario makes one think that we would not be a le to perform even half the intentional a ions agents perform in a day if the agent had to engage in explicit MTT to select each a ion.This is compati le with Mele's (1992) division of intentions into proximal and distal.Distal intentions are intentions to act in the future, for instance, an intention to go to the beach next weekend.A proximal intention, on the other hand, is an intention to act now.
Implicit pro ection plays the relevant role of planning in the production of most intentional a ions, not explicit MTT.Of course, fu l-lown MTT would play a role in some productions of a ion, such as the ones in which the agent explicitly plans what she wi l do; e.g., cases in which she has a distal intention.When an agent has a proximal intention to A, even if it is not salient to the agent that she has such an intention (it is a nonconscious intention),7 she has settled on an a ion plan that she is committed to executing.If implicit pro ection is relevant to the selection of an a ion plan, or if it is the a ion plan itself, then to say that it is part of the intention acquisition/formation is compati le with Mele's (1992) conception of proximal intentions.The question is whether implicit pro ection should be seen as part of the selection of a plan or as the plan itself.It seems that projection does not contribute only to the selection of an a ion plan.The pro ection of the course of a ion selected-the one on which the agent settles-becomes the agent's a ion plan.This is in itself an instance of either implicit pro ection or MTT.Furthermore, by being committed to the plan she is motivated to act accordingly.
If one accepts the ar uments above and that implicit pro ection is enough to make the cost of a course of action affectively salient when one feels the future, then implicit pro ection plays a relevant role for planning and a ion production.As I understand it, feeling the future does not involve any explicit representation; it is a thin experience that inclines the agent toward a course of a ion.The inclination is the affective preference that uides the a ion selection, which is a hypothesis that I consider consistent with cur ent knowledge about a ion selection from intentional maps (Andersen and Bueno, 2002).8

Requirements for moral responsibility
In this section, I wi l discuss the requirement for moral responsibility for a ions and the contribution of implicit pro ection to this requirement.I wi l start by explaining the Filosofi a Unisinos -Unisinos Journal of Philosophy -19(1):89-96, jan/apr 2018 requirement of uidance control.Often the discussion about moral responsibility focuses on whether agents must be free (or whether determinism must be false) in order for agents to be mora ly responsi le.However, this discussion has expanded to include issues such as control.
John Martin Fischer (1994Fischer ( , 2012) ) has ar ued that when moral responsibility is considered dependent on free wi lconsidered as the agent's ability to have done otherwisewhat is in fact being discussed is a certain kind of control that would a low the agent to have done otherwise in the exact same circumstances (in a close possi le world).This kind of control, re ulative control, seems to uarantee that the agent can freely choose what she does; she has different alternatives (a ion courses) open to her, and does otherwise in a close possi le world.Fischer claims that the kind of control that uarantees the possibility of having done otherwise is not necessary for moral responsibility and, based on Frankfurt-style cases, 9 proposes that the agent may be mora ly responsi le even if she could not have done otherwise.I accept Fischer's ar ument that the kind of control the agent needs for moral responsibility focuses only on whether the actual sequence that produces the a ion 10 is moderately responsive (flexi le) to reasons and does not suffer manipulation Fischer's proposal a lows the agent to be mora ly responsi le for her a ion even if determinism is the case, because it focuses on the actual sequence of events that produces the a ion (the production of the a ion), not on alternative possi le sequences.According to Fischer, an agent has uidance control over her a ion if: (1) the actual sequence that produces the a ion is moderately reason-responsive (Fischer, 2012); and (2) the mechanism 11 that produces the a ion is the agent's own mechanism (the agent owns it).Moreover, "the agent's a ion in the actual sequence must be intentional, that is, ap ropriately connected to his reasons" (Fischer and Ravizza, 1998, p. 81).
Criterion (1) means that if the circumstances were moderately different (in a possi le world), then the agent would have a ed differently.Mo erate here is va ue, and Fischer does not offer clear explanation for what is considered moderate.What is clear is that an agent who would only modify her behavior if circumstances were extreme, say, at unpoint, is weakly reason-responsive, while and agent who would modify her behavior in the face of slightly different circumstances (reasons to act) may be considered strongly reason-responsive.From this one can conclude that the agent does not have to be strongly reason-responsive to have uidance control, but she cannot be too weak on her reason-responsiveness in order to have this kind of control.
Reason-responsiveness is considered in the light of the mechanism (process) that produces a ion, which does not have to be an explicit mechanism such as pra ical reasoning.If the mechanism is held fixed across a range of possi le scenarios (possi le worlds) which provide good (sufficient) reasons for the agent to act differently, the agent recognizes these reasons (Fischer, 2012, p. 188), and if the mechanism would produce a different a ion fo the reasons provided, then Fischer considers the mechanism reason-responsive.Active or conscious deliberation is not required for reason-responsiveness; the process may be implicit (Fischer and Ravizza, 1998, p. 64).It is important to note that this is a capacity of the actual sequence mechanism (not a power of the agent), and that mechanisms may vary in degrees of reason-responsiveness (Fischer, 2012, p. 187).
There are two important elements in Fischer's reason-responsiveness: reason recognition and reason rea ivity.Reason recognition requires that the agent should be a le to recognize the reasons there are to act.Reason rea ivity requires that the agent should choose to act in accordance with the reasons she recognizes: "Reasons-rea ivity is the capacity to translate reasons into choices (and then subsequent behavior)" (Fischer, 2012, p. 187).I do not take choice to be necessary; I consider it enough that the mechanism produce the a ion in accordance with the abovementioned reasons to act (2012, p. 187).
According to Fischer and Ravizza, reason receptivity must be strong, meaning that "the agent would recognize what reasons there are, given that the actual kind of mechanism operates" (1998, p. 69).In fact, the mechanism should not be evaluated case by case, what Fischer and Ravizza (1998, p. 71) have in mind is that the mechanism should exhibit a pattern of (actual and hypothetical) reason recognition minima ly grounded in reality.
Criterion (2) has to do with cases of manipulation.It does not concede that the agent has uidance control in cases in which the mechanism that produces the a ion is manipulated, which may be the case in sci-fi scenarios that a low for crafty neuroscientists, alien chips, demons, and the like.The criterion, however, also emphasizes that the actual sequence 9 I consider Frankfurt-style cases the ones in which the agent could not have done otherwise (Frankfurt, 1969).There are several examples of such cases in the literature, which may be roughly simplified as cases in which the agent has a morally relevant choice: (1) to do A, or (2) to do B. In the imaginary scenario there is an external manipulator, for instance a demon, that would block the neural mechanisms that would allow the agent to choose doing A if the agent is ever inclined to do so.Nevertheless, the agent is not inclined at all to do A, and just does B. The agent in question could not have done otherwise, but she still seems to be morally responsible for her action, because the demon never interfered in the production of her action.Several complications can be added to this kind of scenario, but I believe that enough has been said to clarify the argument that moral responsibility does not require the ability to do otherwise.For a discussion about Frankfurt cases and kinds of freedom see Fischer (2008). 10When referring to the actual sequence, Fischer states that it "issues in action".However, considering that I accept a causal explanation of action, I will use the expression produces action. 11By mechanism, Fischer means process (Fischer, 2012, p. 10).
that produces the a ion results from the agent's own mechanisms.Together, satisfying the two criteria a ds up to the agent having uidance control over her a ion, which means she is mora ly responsi le for her a ion.
The conditions for moral responsibility focus on the production of the agent's a ion, on the actual sequence.In other words, whether the process that produces the a ion suffered any wor isome form of manipulation, whether the process is responsive to reasons.Any consideration about de cognitive capacities of the agent that make these possi le as a condition for moral responsibility would fit in this part of the discussion.This is where one may wonder about MTT.If my claims about implicit pro ection in the previous section are accepted, then one may conclude that the agent's ability to feel the future, a pro ection that has a phenomenal component (but does not involve any explicit representation) helps explain Fischer's uidance control.In fact, it seems relevant for both criteria 1 and 2. I wi l discuss these in the fo lowing.
It is reasona le that the mechanisms that produce a ion depend on implicit pro ection, such as cases in which deliberation is not explicit.When an agent settles on a course of a ion, she relies on implicit pro ection, at least, to estimate possi le courses of a ion, which provides her with reasons to act.This shows its contribution to criterion 1; pro ection aids reason-responsiveness to different possi le future contexts-which is in accordance with the weight Fischer puts on counterfactual sets of reasons.Of course, not every step of this process has to be explicit to the agent; for instance, when I see traffic on my favorite route home, I feel a little annoyed and I immediately take the longer way, which is usua ly free of traffic.I do not have to explicitly represent the pros and cons of each route, nor do I explicitly deliberate before I settle on changing routes.The process seems to hap en much faster, based on previous times when I was stuck in traffic on that route, which were ir itating and unpleasant experiences.On the other hand, I could have explicitly considered the pros and cons remembering the times that I took that route despite the traffic, and I would have reca led the experienced of being stuck in traffic, according to the MTT theory, which would proba ly have made me turn away from that route as fast as I could.
The mechanism does not have to be pra ical reasoning.In fact, Fischer and Ravizza (1998, p. 86) defend that their account ap lies to nonreflective mechanisms just as we l as long as they are reason-responsive.In their view, the same kind of mechanism can produce different a ions in the face of different sufficient reasons to act that the agent recognizes.Even a pattern of reason recognition can be compati le with nonreflective mechanisms.The recognition of reasons does not mean that the agent has to explicitly weight her reasons or ask herself what to do.Fischer and Ravizza consider that it may be the case that a ions produced by nonreflective mechanisms are produced by the agent's traits, which are not reason-responsive; nevertheless, in the history of the a ion, uidance control can be found in the formation, retention, or expression of the trait.The latter is sufficient to consider the mechanism reason-responsive.
There is, however, more to implicit pro ection's contribution to uidance control.This leads to criterion (2).In Fischer's account, the mechanism that produces the a ion is the agent's own mechanism if she has some beliefs about her agency's effects in the world.In a sense, this means that the agent takes responsibility for a ing as she does (see Fischer, 2012, p. 187).According to the MTT theory, it is not enough for the agent to semantica ly remember the past or think about the future; if the agent cannot think of these in an autobiographical context, then it wi l fail to have affective significance to her (Ger ans and Kennett, 2010, p. 598).This is the case because the autobiographical context gives meaning to the agent's decisions and plans, which make it he plan.It is not enough that one course of a ion is selected among possi le alternative courses of a ion; the agent needs to stick to it.The phenomenal relation to her decisions and plans bound her to them.
Planning, taken as this kind of projection, commits the agent to a ing according to the plan when the time comes, because she has the experience of inhabiting the future she plans, therefore, she feels connected to her plan and settled on executing it; it is he plan (Ger ans and Kennett, 2010, p. 604).The plan is made taking herself as tempora ly extended into consideration (2010, p. 605), as we l as the affectively positive valence of what the plan wi l produce for her.Boyer (2008) considers that the key element to explain MTT's adaptive value is MTT's contribution to decision-making in planning, for it a lows the agent to restrain from immediate rewards to accomplish (possi ly through cooperation with others) future rewards that are worth waiting for by making the affective valence of possi le future situations immediately present to the agent.The affectively positive valence would motivate the agent to stick to a distal plan, but also an affectively negative valence of a future consequence from immediate reward may make the agent stick to a plan that includes refraining from the immediate reward.
The emphasis placed on the agent's connection to her plan helps account for what it means for the mechanisms that produce the a ion to be her own.This shows why cases of manipulation in which the manipulator (neuroscientists, demon, etc.) tampers with or renders ineffective the agent's a ion selection process-which can be implicit pro ection, explicit decision-making, or planning-take away the agent's uidance control. 12side from the obvious fact that someone else interfering with the process that produces a ion would eliminate uidance control, I claim that manipulation of this kind would take away the agent's feeling of ownership of her plan, her commitment to act according to it, and her feeling of ownership of her a ion.These are part of MTT' s, and implicit pro ection' s phenomenology, so if the agent does not fo low her plan, the production of the a ion is not accompanied by the phenomenal feeling of ownership inherited from the a ion plan.If the agent does not simulate her future (implicitly or explicitly) in order to select an a ion course, she does not own the a ion plan, nor feels that it is her plan, and she can only be artificia ly committed to a ing, i.e., through manipulation.Nevertheless, in the latter case, she would not have uidance control over her a ion, nor can she be held mora ly responsi le for her a ion, according to the view accepted here.
If a l of the above is cor ect, then even a modest theory of moral responsibility benefits from, at least, implicit pro ection, for it has been explained above that implicit pro ection helps explain and contribute to Fischer's requirements for moral responsibility.It also contributes to the motivational a ect of a ion production.Therefore, at least implicit pro ection is relevant to moral responsibility, i.e. a form of pro ection is part of the agent's production of a ion whenever she has uidance control over her a ion.

Possible objections
It could be claimed that I am a ding too many criteria, thus, setting the bar too high for moral responsibility.But this does not seem to be a strong objection, because I am not a ding any requirements for moral responsibility.I have simply claimed that if one accepts that uidance control is the condition for moral responsibility, then one may need to accept that MTT or implicit pro ection plays a relevant role in this kind of control.
Another objection may be that theories that require the agent to have alternative possibilities (the ability to do otherwise) in order to have moral responsibility would also encompass MTT or implicit pro ection; therefore, there is no reason, from MTT alone, to focus on uidance control.This may be true.If the agent deliberates or chooses from alternative possibilities, it could be said that the same processes of MTT, or implicit pro ection, are in some sense relevant to her choice.
The affective valence of the simulated alternatives, however, would have to be different.Consider that the agent projects herself in a future where she decides to do A instead of B, and the projection has an affectively positive valence, whereas when she projects herself in a future where she does B, it has an affectively negative valence.This explains why she does A. Now, in a possi le world where the agent does B instead of A it seems that she would have to have the op osite affective responses and a different psychological constitution that would ground such difference.If this is cor ect, the agent would not in fact act otherwise in a close possi le world, only in a far enough one to a low for so many differences from the actual world.Therefore, MTT may pose more cha lenges than solutions to alternative possibility theories.

Conclusion
My aim has been to show that MTT and implicit pro ection help shed light on the requirements for moral responsibility in human agents, if one accepts that uidance control is the necessary and sufficient requirement for moral responsibility.I have claimed that implicit pro ection and sometimes MTT in the form of planning play a relevant role in the actual sequence that produces the agent's action, therefore contributing to the agent's uidance control and moral responsibility.