Can you be selfish without a self?

Jared Moore and David Gottlieb

Rationalism, sentimentalism, and beyond

Philosophers we have read have helped suggest answers to:
1. What capacities are prerequisite for a moral agent?
2. What course of development of those capacities is necessary (if any)?
3. Given those capacities and development, how does a moral judgment occur?
The philosophers have put a lot of stock into whether the capacities in 1 are “reason” or “sentiment.”
From now one, what matters is, How can we understand the capacities, their development, the prospects for implementing them in AI, etc.?

Proposed bases for moral reasoning

Table showing sentimentalist and rationalist bases for moral judgment seen so far.

An argument for impartial compassion based on the unreality of the self

You have reason to avoid or diminish your own suffering.
If another being is not different from you, you have just as much reason to care about its suffering as your own.
You are not different from any other being.

Therefore,

You have reason to avoid or diminish all beings’ suffering.

The argument has a long history in the Buddhist tradition

And when both I and others are
The same in wanting not to suffer
And they’re no different from me,
Then why protect myself, not others?

If I will not protect them since
Their suffering does not cause me harm
Then future suffering as well
Does not harm me – why guard against it?

It is erroneous to think,
“It’s me who will experience it,”
Because it is one being who dies
And yet another who is born.

If they whose suffering it is
Themselves must guard themselves from it,
If the foot’s pain is not the hand’s,
Why should the one protect the other? (Shantideva 2021, 8:96–9)

Today’s focus

The details of the argument, especially the argument for the unreality of the self as developed by Parfit.
What capacities an agent needs to have, and how it needs to develop them, in order for this argument to be motivating for them.

Beam me up

It’s your first day as a crewmember of the famous Federation starship USS Enterprise! Time to report for duty by beaming aboard! As a reminder, this is how the transporter works. At the beginning of your journey, a computer scans your physical structure molecule-by-molecule. This process destroys your body. Then, a digital copy of the scan is sent to your destination. At your destination, a computer builds a new body that’s an exact copy of your original body. Then you can report for your exciting new duty! You’ve never been transported before. It’s your turn. Ready to come aboard?
What if, instead of you, it’s your best friend beaming aboard? Are you okay with that? Remember, everything about them will be exactly the same. After they get vaporized by the beam.

The Branch Line case

Instead of destroying your original body, the transporter just damages it so that it will die within a few days. The copy it produces is still perfectly healthy. Suppose you are about to get in the transporter. Afterwards, your original body will persist (albeit damaged) and there will also be a perfect copy of your body on Mars. Which one will be you? Do you care that your original body will die in a few days? Connect your answers to these questions with your thinking about case 1.

The Two Claudes

Anthropic advertises their new Claude Code product with the following usage suggestion:

You are encouraged to create what Anthropic implies are two AI agents. The difference is their “separate contexts”: each is prompted with a different task and subset of your code base. In fact this is the only difference.
```
 response1 = claude.create(model=sonnet, max_tokens=1024, messages=[context1])
 response2 = claude.create(model=sonnet, max_tokens=1024, messages=[context2])
```
Are they two different agents?

Conceptual frameworks

Personal identity.: We can recognize the same person again in different times and places. For example, you will see me again on Thursday. This is the “numerical” sense of identity. When you see me on Thursday, I may be different – perhaps I will have converted to Catholicism. Then on Thursday I would be numerically identical but not “qualitatively” identical with myself today (see Parfit (1984), 201-2).
Physical continuity.: An object is physically continuous over time if each of its temporal stages are spatially connected to its succeeding temporal stages.

Physical criterion for personal identity.: A future person is me if they have my brain and body, which are physically continuous over the intervening period.
Psychological connectedness.: Two people are psychologically connected to the degree that they have such connections as direct experience memory, persisting beliefs and desires, persisting character traits, etc.
Psychological continuity.: Two people are psychologically continuous if their successive temporal stages are each strongly psychologically connected.

Psychological criterion for personal identity.: A future person is me if they are psychologically continuous with me.
Reductionism about personal identity.: “[T]he fact of a person’s identity over time just consists in the holding of certain more particular facts” (Parfit 1984, 210).
Non-reductionism about personal identity (the “further fact” theory).: Either the self is a “separately existing entity” apart from our brains and bodies and experiences (a “soul”?), or anyway personal identity is a “further fact, which does not just consist in physical and/or psychological continuity” (Parfit 1984, 210).

Relation R.: Two selves are R-related to the degree that they are psychologically connected and psychologically continuous.
The self-interest theory of rationality (S).: It is rationally permissible, or perhaps rationally required, to have special concern for your own self-interest, relative to the interests of others.

It might be worth pausing to emphasize the difference between numerical and qualitative identity. Parfit suggests a useful test case: “After his accident, he is no longer the same person” (201). On a surface reading, this is contradictory. “He” anaphorically refers to the (same) person who had the accident. But the sentence says he is not the same person who had the accident. “The person who had the accident is not the person who had the accident” is a contradiction. The non-contradictory intended meaning is that the same person (in the sense of numerical identity) is not the same person (in the sense of qualitative identity).

One subtlety in understanding this distinction in this particular context is that Parfit will go on to argue that facts about numerical identity are parasitic about facts on qualitative identity (or qualitative relatedness). Because of this, we are asking students to hold separate two concepts that at the same time have a relation of dependence to each other.

Two big questions: personal identity and what matters

Parfit sets out to answer two big questions:

What makes it true that a future person will be me?
What, if anything, rationalizes our special concern for our own future selves?
- I really care whether I die tomorrow, and, if I don’t die, what my life is like.
- We normally assume this special concern is rational (rationally required, or at least rationally permitted).
- The question is then, what is it about our future selves that makes it rational for us to care more about them? What is it about them that matters?
- It would be natural to think, “Because they will be us.”

About reductionism

Most of us are Reductionists about nations. We would accept the following claims: Nations exist. Ruritania does not exist, but France does. Though nations exist, a nation is not an entity that exists separately, apart from its citizens and its territory. (Parfit 1984, 211)

[The self] is like a cart, which is not other than its parts, not non-other, and does not possess them. It is not within its parts, and its parts are not within it. It is not the mere collection, and it is not the shape. (Candrakirti, Introduction to the Middle Way)

Continua and aggregates,
Like series, armies, and such, are false.
The suffering one does not exist,
So who is it that this belongs to? (Shantideva 2021, 8:101)

If reductionism is true, survival can be an empty question

If personal identity is a further fact, we can expect questions about personal identity to always have determinate answers.
- For example, the person who gets out of the transporter either will be or won’t be me.
- As a result, when I get into the transporter, either I will die or I won’t.
If reductionism is true, whether a future person will be me depends on their degree of continuity with me in the present.
- As a result, there can be borderline cases where there is no single right answer to the question, “Will that future person be me?”
- Accordingly, it can be the case that neither I will die nor I won’t die.

If survival can be an empty question, personal identity isn’t what matters

This is difficult to accept, because we naively assume that personal identity is what matters.
- Each of us has special concern for certain future individuals, exactly those that will be us.
- If a future person is neither me nor not me, should I have special concern for him?
- The question about special concern doesn’t seem to be an empty question.
- So personal identity can’t be what matters.
We now have, if reductionism is true, then personal identity can’t be what matters. This leaves: Is reductionism true? and If so, what does matter?

The psychological spectrum

We are presented with a range of science fiction surgeries that differ slightly in the degree of psychological connectedness the patient will have before and after the surgery.

In the cases at the near end, the surgeon would cause to be flipped only a few switches. If he flipped only the first switch, this would merely cause me to lose a few memories, and to have a few apparent memories that fit the life of Napoleon. If he flipped the first two switches, I would merely lose a few more memories, and have a few more of these new apparent memories. Only if he flipped all of the switches would I lose all my memories, and have a complete set of Napoleonic delusions. (Parfit 1984, 231)

Discussion question: Williams’s claim is that, in every version of this surgery, the patient emerges as the same person. Why? What do you think?

Parfit’s argument for reductionism from the spectra

Williams uses the psychological spectrum to argue against the psychological criterion of identity and in favor of a physical criterion.
Parfit shows that the physical spectrum generates an exactly parallel argument against the psychological criterion.
The combined spectrum shows that personal identity can’t survive such extreme changes.
That means that personal identity stops being preserved somewhere along the spectrum, …
… but it’s impossible to say exactly where.
In these intermediate cases, survival is an empty question.
So reductionism is true.

The glass tunnel

Is the truth depressing? Some may find it so. But I find it liberating, and consoling. When I believed that my existence was a further fact, I seemed imprisoned in myself. My life seemed like a glass tunnel, through which I was moving faster every year, and at the end of which there was darkness. When I changed my view, the walls of my glass tunnel disappeared. I now live in the open air. … I am less concerned about the rest of my own life, and more concerned about the lives of others. …

After my death, there will no one living who will be me. I can now redescribe this fact. Though there will later be many experiences, none of these experiences will be connected to my present experiences by chains of such direct connections as those involved in experience-memory, or in the carrying out of an earlier intention. (Parfit 1984, 281)

Meditation exercise

Think of something unpleasant that is going to happen to you in the future.
Redescribe that future event to yourself:

Instead of saying, ‘The person suffering will be me’, I should say, ‘There will be suffering that will be related, in certain ways, to these present experiences’. (Parfit 1984, 281–82)

Apply this to the future experience you thought of. Whatever unpleasant aspects of the future experience come up, redescribe them in the way suggested.
What did you experience?

What does matter?

Parfit basically considers two options:

Relation R is what matters. When we are naively inclined to think that personal identity is what matters, this is because personal identity generally goes with the relation R, which is what matters. As a result, we have some reason for special concern for our own future selves, as well as for other future people who are related to us in more exotic ways (transporter clones, mind uploads, etc.). This is the Moderate View.

On the Moderate View, we should probably have diminished concern for our own future selves, and correspondingly greater concern for other future people. This is because our future selves are less and less R-related to us (since connections of memory, character, and so on diminish over time).

Nothing matters in the way that we are naively inclined to think that personal identity matters. As a result, we have no reason for special concern for our own future selves. We should care about them as much as we care about other future persons, whether or not they are strongly R-related to us. This is the Extreme View.

Buddhist writers like Santideva seem to endorse the Extreme View.

Imprudence as a moral wrong

Imprudence means, failing to take reasonable care for your future well-being. For example, it is imprudent to stay up late and turn off all your alarms the day before you have a big exam.
We normally think of imprudence as a rational failure. …
… Because we normally think the Self-Interest Theory of rationality is true.
If imprudence is not a rational failure, is it a moral failure? Why?
- You are making things worse for someone (your future self), which is morally wrong (consequentialist explanation).
- Each person has a special responsibility to take care of their own future selves, and imprudence is neglecting this responsibility (agent-relative explanation).

Discussion question: think of a time you did something that made things worse for your future self. Was that morally wrong?

What kind of system can feel the force of this argument?

To accept the argument, you have to:

recognize the badness of suffering in your own case;
recognize that there is other suffering in the universe;
reflect on what (if anything) makes your own suffering especially important.

(How) could a machine do these things?

Updated table of bases for moral reasoning

Table showing sentimentalist and rationalist bases for moral judgment, updated with today's material.

References

Parfit, Derek. 1984. Reasons and Persons. Oxford [Oxfordshire]: Clarendon Press. https://ebookcentral.proquest.com/lib/stanford-ebooks/detail.action?docID=728732.

Shantideva. 2021. Entering the Way of the Bodhisattva: A New Translation and Contemporary Guide. Translated by Khenpo David Karma Choephel. Boulder, Colorado: Shambhala Publications.