Mathematical models will never simulate human behavior precisely.

Different models show different results. So, if some model provides some precise conclusions about some community, it is not necessarily the same for the others.

People's insights, morals, freedom, and responsibility are very complicated to be captured and recorded in exact equations or tables. Some approximated partial results are accepted for making decisions about some phenomenon.

No.

This, at any rate, is the conclusion. For the premises (and thus the argument), see

Article The Mandatory Ontology of Robot Responsibility

Article Bridging the Responsibility Gap in Automated Warfare

Mathematical models will never simulate human behavior precisely.

Different models show different results. So, if some model provides some precise conclusions about some community, it is not necessarily the same for the others.

People's insights, morals, freedom, and responsibility are very complicated to be captured and recorded in exact equations or tables. Some approximated partial results are accepted for making decisions about some phenomenon.

Resolve questions of human freedom and responsibility? Perhaps to a limited extent. For example, the more nuanced treatments of free will and determinism might highlight grey areas in which they are not mutually exclusive. However, AI systems would have to transcend their hardware and software limitations and become pukka moral agents like humans, for moral decision-making and responsibility. Perhaps the incorporation of biological components (biochips and other wetware) may eventually achieve something like that by providing the right sort of causal substrate.

Following

This is certainly a very complex debate, for which there are no easy answers. I, myself, would argue that machines and artificial intelligences will never quite be able to fully represent the paradox of freedom and determinism in human beings, in part because human beings can't necessarily be entirely explained by completely causal models. Human beings have the unique experience of being a creature whose actions have many causations behind them, and to make matters even more complex, humans are also uniquely able to transcend all of these causations, making us more than mere cause-and-effect based beings (although I likely wouldn't go as far as to argue that cause and effect play no role in human action and decision-making.) As Issam Kaddoura and Karl Pfeifer mention, machines and AI's may be able to mimic certain aspects of humanity, but as of now not all aspects of human experience. I would go even further than them and say that the one thing about human existence that such creations will never be able to imitate is that humans can be understood phenomenologically, or as having a subjective, lived experience that is deeply intertwined with their external, objective world, while I do not believe that such a fundamental aspect of humanity can be replicated in machines or AI's (they may be able to imitate it, even give the impression that there are more subjective experiences occurring beneath the surface of their more mechanistic functioning, but I don't think they will ever quite be able to truly have such experiences.) Of course, many people are likely to disagree with me on this point, and I will not act as if I have all of the answers—I may very well be wrong. But my conviction is that while machines and AI's will always be able to be reduced to certain mechanisms and programmings, humans, despite the variety of people who believe they can also be reduced to such forms of functioning, will always be too complex to be simplified down to such a mechanistic model.

Advaita Vedanta model (one of the major idealist school in Indian philosophy) can solve the problem.

Niladri Das Would it be possible to give a brief summary of this school of thought and the answers it offers to this problem? I know that no philosophy can be done full justice with such a short summary, but I, for one, am interested in your take on how it relates to this question, as I'm sure a few others here are as well.

In brief, Advaita philosophy basically rejects the notion of causality since causality necessarily involves a subject-object distinction which is unacceptable to Advaita (non-dualism).

Domenic Marbaniang Thank you for that! It sounds like the Advaita Vedanta model has some commonalities with the Western philosophical school of Phenomenology, which also seeks to transcend the subject-object distinction and embrace a non-dualistic approach to thinking. Phenomenology does not go quite as far as Advaita in rejecting causality altogether, but it does attempt to suspend all assumptions of causality initially in order to gain a more holistic understanding of a phenomenon first (and holism largely relies on non-dualism.) It would sound that Phenomenology (a school of thought that I've found myself highly influenced by) and Advaita Vedanta have a good possibility of dialogue set up for this issue of causality, and their converging assumptions may allow them to be closely allied in their arguments against the use of pure causal models.

Domenic Marbaniang and Christopher Zieske... According to Advaita Vedanta school, there are three grades of reality, 1) momentary level: lower level (illusory state like rope-snake illusion etc.), 2) Phenomenal level: middle portion ( this world) and 3) Transcendental level: highest state ( Ultimate where you can find yourself as the identical being of Brahman, it means 'you are Brahman' ). Brahman is within us... due to ignorance we are covered by imposition like the body is mine. In this case, the body is matter and soul is a spirit but due to ignorance we think body and soul are identical like ' my body'. each and every individual level is real until our self-realisation or transcendental divinity. In this manner, the phenomenal world is always real from the phenomenal perspective but when being is in the transcendental level then the ignorance will be gone and all phenomenal beings are false. In this transcendental level, there is no causal connection. This is the metaphysical structure.

Now I would say that according to this system, in this phenomenal level there are some chronological processes to deal with society. If we follow these, then we can achieve our goal and I think modernity comes new feet with old shoes.

Niladri Das Thank you for your elaboration on this discussion! I will continue to discuss the "dialogue" between Advaita Vedanta and Phenomenology that I mentioned above, as I still see many opportunities for such a dialogue here.

Where Advaita Vedanta and Phenomenology are in agreement: Phenomenology would agree that there are multiple "modes" of reality, each differing from the others in some way despite being related to each of the others. Phenomenology would also agree that reality can be misleading, as we can "cover over" reality upon our first perception of it and thus only understand one aspect of reality as it presents itself to us rather than the whole reality (see Martin Heidegger's "Being and Time.") This appears to be essentially what happens at the "momentary" level of reality as described by the Vedanta model, as it involves illusory perceptions. Phenomenology would also agree that there is the phenomenal level of reality (the world that is immediately accessible to us) and a transcendent level of hidden meanings, which can only be accessed by turning our capacities for reflection upon our own experiences of the phenomenal world in order to discover what we may have initially missed.

Where the two schools may disagree: Phenomenology would argue that the phenomenal world is not false, and that the transcendental world is, indeed, simply a further elaborated perception of the phenomenal world. Phenomenology would put forward the visualization of the "transcendental" level of reality as being a look at the phenomenal world from above it and outside of it, rather than within it, as opposed to Vedanta's visualization of transcendent consciousness as looking away from the phenomenal world (it could also be referred to as the "meta-phenomenal" level of reality.) Likewise, Phenomenology would disagree with some of the points made by Advaita Vedanta about transcendence, such as the idea that the body and soul are separate—indeed, to a phenomenologist, body and soul are identical, and one's body is one's own (although not in the sense that the soul exists in some "part" of the body, like may other philosophical schools believe.) The soul must always be embodied in some way in order to exist (this does't inherently criticize the idea of life after death, however, as it can be put forth that if there is life after death, then the soul simply takes on another embodiment—it is up to the individual Phenomenologist where they want to take their conception of the "embodied spirit.")

Would anybody have anything they'd like to contribute to this discussion?

Christopher Zieske , considering that Advaita does admit causality at the phenomenal level (it distinguishes between various states of consciousness: dream, dreamless sleep, wakefulness, turiya or pure consciousness), it will still reject the proposal that a machine can make ethical decisions for the following reasons: (i) the body is viewed as analogous to a garment that the spirit changes upon death; therefore, physical machines are not at par with the spirit (ii) the phenomenal spirit or self is the volitional and moral agent, not the body.

The Buddhist view, on the other hand, sees personality as emerging from the aggregation of form, sensations, perceptions, mental activity, and consciousness contradicting belief in a permanent self/spirit (doctrine of anatta). It might seem to be closer to the "volitional" machine-assemblage concept, perhaps. However, the material-form is just one of the five aggregates of phenomenal reality.

Domenic Marbaniang I would agree with Advaita's assertion that a machine can make ethical decisions, but where I think the difference between the thinking of Advaita Vedanta and Phenomenological thinking occurs is in the discussion over whether or not the body can be seen as a machine. Where Phenomenology disagrees with the status quo of much philosophy worldwide is in its assertion that the body is not a machine, and cannot be understand in purely mechanistic terms. Because Phenomenology sees the body and soul as being identical, and as being united as one and the same through human existence, it would assert that to attempt to understand the body as a mere machine carrying out tasks which it was physically designed to do would miss the fact that the body is a fundamental and indispensable part of the volitional and moral agent, which, to a phenomenologist, is not just the spirit alone, but the person as a unified totality. In this sense it would seem that Phenomenology would agree somewhat more with the Buddhist perspective, although it seems that the Buddhist view (at least from your description of it) still sees the body in somewhat of a mechanistic light. It is certainly true that the material-form is only one aspect of phenomenal reality; however, from a Phenomenological perspective, none of the other aspects of phenomenal reality (in the sense used by Phenomenologists themselves) can exist without the material form, just as the material form cannot exist without any of the other aspects of phenomenal reality, as for a Phenomenologist, any reality must be understood as a unified totality with which no one aspect can be dispensed.

I hope you have found this conversation as stimulating as I have.

Domenic Marbaniang There was a typo in my last answer. I had intended to write that machine's cannot make ethical decisions, but I had mistakenly written that they can. I hope that this does not cloud the meaning of my statements too much.

Christopher Zieske , yes it's interesting to have a wider scope of approaches to the issue. Thanks!

Domenic Marbaniang Of course! Certainly, I think that it is always good to set up dialogues such as these between differing schools of thought, as each one can provide a contribution of its own and thus allow for genuine further understanding of the issue (as opposed to when one school of thought wins out over others and it seems to simply be blindly accepted.) I am also glad that we were able to set up a dialogue between Eastern and Western schools of thought on this subject, as I simply think that there hasn't been enough discussion between Eastern and Western philosophy so far (making each miss out on a point of view that could help to develop a more complete answer to difficult questions.)

There seem several - very interesting - questions tied together here, which makes sense since the topic you're asking about is itself complex. I read you asking at least the following questions:

(1) Is it possible to construct accurate causal models of human behavior?

(2) Assuming the answer to (1) is "yes", is it possible to construct accurate causal models of human behavior sensitive to human freedom?

(3) Assuming the answers to (1) and (2) are "yes", can such causal models provide plausible moral guidance for humans?

With respect to (1), I'm inclined to say "yes". In fact, I think much progress has been made on this already in behavioral economics, philosophy, and psychology. As an example of progress in the last area, I recommend Lance Rips's "Two Causal Theories of Counterfactual Conditionals" as a start.

With respect to (2), I think the answer to this depends on what's meant by "freedom". Philosophers often divide related discussions up by first defining:

• Free Will - Some actions are free
• Determinism - Every event has a prior cause
• Incompatibilism - If determinism is true, then no action is free

Clearly, these can't all be true at the same time. Three common responses:

• Hard Determinism - Determinism and Incompatibilism are true, so no action is free
• Libertarianism - Incompatibilism and Free Will are true, so Determinism is false and some actions have no prior cause, i.e. are free
• Soft Determinism - Determinism and Free Will are true, so Incompatibilism is false, and we sometimes act free despite our actions being caused by prior events

I think a plausible way of understanding (2) amounts to it being a disguised version of (1), specified with commitment to one of these responses. For example, if you accept Hard Determinism, and "freedom" in (2) amounts to you're wondering if we can develop causal models that accurately predict human behavior, then I think this question is actually a disguised form of (1), and the answer seems "yes" since events can be predicted. I'm inclined to say the same if you accept Soft Determinism. On the other hand, if you accept Libertarianism and are wondering if we can develop causal models that accurately predict human behavior, then it's not so obvious we can predict human behavior, since it's uncaused. Assuming I've accurately represented a question you asked, which - if any - of these positions do you have in mind?

With respect to (3), let's assume we've a complete accurate causal model of human behavior capable of excellent prediction. Can this provide moral guidance? Perhaps, but we should be careful not to fall into the trap of thinking that simply because humans behave in a certain way that they should. A causal model that accurately predicts that nearly everyone will litter itself doesn't generate a prescription that people should litter. I don't actually think you're in danger of making this error, I just want to be careful here is all. That aside, I'm inclined to say that an accurate causal model as described would likely be representative enough of our values to illuminate what each of us believes we should do, even if we don't do it. For example, such a model might predict that we litter, but it would presumably also predict that most of us would agree this is not something we should do. That combination of prediction plus evaluation might be components of moral guidance you're looking for.

Thanks for this excellent question.

John Beverley Thank you so much for your contribution to this discussion! Indeed, this question does seem to have many facets to it. I know that your intention was likely to get the opinions of Domenic Marbaniang on these questions, but if you don't mind, I'd also like to provide my own takes on these questions and get your feedback on my responses.

1. I certainly won't disagree with you that causal models of human behavior are possible to a limited extent, but I might limit the statement by saying that a purely causal model is not possible. Causal models could, perhaps, have a wide range of use in predicting human behavior, but I would assert that it is unlikely that any causal model capable of predicting exactly what a person will do in any given situation is just outside of our reach, and so we must approach causal models with caution and the awareness that they only account for a limited range of human experience.

2. Not only would I argue that a causal model for human behavior can be sensitive to human freedom, but I would got further and say that the ideal causal model would be sensitive to human freedom, thus acknowledging its own limitations in predicting behavior and explicating the fact that the causal model is only one factor of many that play a role in human behavior.

3. I would essentially agree with you on this question. A causal model, since the model itself is not a human being with the distinct capacity to reflect on its own action and the implications of them, can only really make descriptive statements rather than explicitly normative statements. Descriptive statements can be used to make implicitly normative statements (a fact that essentially formed the basis of psychologist Abraham Maslow's entire writing career), but the normatively of these statements must come from a being capable of reflecting on morality and ethics, i.e. a human being.*

I look forward to hearing what everyone has to say on this topic.

*I know that some species of animals have been found to show some capabilities for moral introspection as well, but in these cases, I don't think that we can be completely certain that this capacity is as deeply developed as it is in humans—it may be, but we are not certain enough of it to take it as a given.

Thanks so much Christopher Zieske for your additions. It occurs to me - but hadn't before you answered - that Newcomb's Paradox may provide a way to extend the issue in (3) to non-moral guidance, insofar as we have an ideal causal model. In the event you're unfamiliar:

A researcher has created an ideal causal model (or something close enough to it), that has never failed to accurately predict choices in the following scenario. There are two boxes before you, A and B, and you've the choice of taking only A or both A and B. You know there's $1,000 in B. Additionally, you know that if the researcher believes - based on the causal model - that you will take box A only, then he has put $1,000,000 in A. Otherwise, he left box A empty. Which do you choose, box A or both A and B?

The tension emerges because the following two decision-theoretic principles seem plausible:

• Given the evidence, choose so to bring out your desired outcome
• Only act if your action will change the outcome

The first suggests you should choose only box A, since you can expect A to be empty if you choose it based on your evidence. The second, however, suggests choosing both since choosing A won't affect the outcome.

In other words, assuming we've an ideal causal model which reliably predicts our behavior, it seems at best we could only be provided plausible options based on expected utility, but the choice between them seems to depend on something like - perhaps - weighing utilities in ways causal models don't typically do in decision-theory as far as I understand. I've (clearly) not thought enough about the relationship here, but I appreciate you spurring the thought, and I thought it'd be worthwhile to share.

John Beverley Thank you so much for bringing this up! It sounds, if I'm reading correctly, as if Newcomb's Paradox essentially states that even if an ideal (or near-ideal) model were possible, it still wouldn't be able to fully predict behavior, because it wouldn't be able to account for certain variables in the way humans measure utility, especially if they are aware that their actions are being predicted—would you say that this is accurate?

John Beverley, thanks for tackling the root issues involved and the discussion at length. With regard to "which - if any - of these positions do you have in mind?", ignoring Laplace's demon (strict/hard determinism), we are left with the other two. But, I might go for soft determinism for the sake of discussion. Of course, we will need to add other variables that include moral cognates in the equation.

Suppose soft determinism..., and we admit either (a) deontological or (b) teleological structure as the moral framework for a machine's decision-making, perhaps (a) would be plausibly simpler as it doesn't need to possess an infinity of counterfactual info required for (b) for the sake of perfection. But, that would still beg the question for the issue of constructing a morally accurate thinking machine transcends the choice between (a) or (b). In other words, one has to construct a machine that first decides what morality is before it can make accurate moral decisions. That would imply creating a machine that is self- determining. But, then how can one be sure that this machine would make accurate predictions and decisions? Yes, this certainly involves several paradoxical results....

Many thanks also to Christopher Zieske for adding to the discussion.

As scholars one of our tasks is to link our fields of study with ontological constructs. Realist and constructivist frameworks are useful in that - while the first acknowledges real mechanisms in society, such as culture, social structures and human agency, the second values the inner and psychological interpretations of these real phenomena. Any explanation or analysis of "freedom" should therefore consider human subjectivities...as emerging from the mind and in relationship to our navigation of the real world.

Article Friends or foes? A theoretical approach towards constructivi...

@Oscar Oliver Eybers, I would think that such an hypothetical AI machine will have to be capable of becoming aware of the various contexts. However, will it be capable of the subjective aspects? That leads to a wholly important debate ... But, if I can limit the scope, assuming a deontological/legal framework only, can an AI replace human judges in judicial courts...? Note that we have demoted them from lawmakers to judges...

Oscar Oliver Eybers I had the pleasure of reading the article you had shared. To explicate its relevance to the present discussion, allow me to give my interpretations on what each of the ontologies you describe in the article would likely say on the matter of whether causal models or machines can solve issues of human freedom and moral choice. Please feel free to refute either of these if you feel that they don't quite fit the ontologies' basic assertions.

1. Constructivism, with its focus on human subjective and intersubjective experience as the primary defining factor of reality and being, would seemingly argue that causal models and machines can not quite solve such issues, as they would fail to fully comprehend these fundamentally subjective and psychological phenomena. To borrow terms from Dilthey, such models would overemphasize the erklären mode of being (understanding reality objectively or from the outside of the subject) at the expense of the verstehen mode (understanding being and reality from within the subject's own experiences.)

2. Realism, with its emphasis on reality as an extraconscious entity independent of human subjective experience, would seem to argue that such models are capable of solving issues of freedom and moral choice, as they would claim that such models are better capable of detecting the aspects of these issues that originate from outside of human experience—perhaps even better at doing so than humans themselves. Returning to Dilthey's terms, realist ontologies seem to emphasize the erklären mode of being over the verstehen.

In your article, you argue that despite the initially contradictory assertions of these ontologies, they do not necessarily need to be antithetical to each other. I wholeheartedly agree with this statement, but attempting to combine perspectives such as these that have a history of opposition in such a way that each perspective is equally represented is certainly no easy task, even if possible. Thus, I think that the most important question that this raises is, how can we work out an ontological framework that equally synthesizes both realist and constructivist ideas? And, specifically for this discussion, how can we combine realist and constructivist views on the relationship between causal models and artificial intelligences and concerns over human freedom and moral decisions?

Would anyone want to give their take on this issue?