Explaining to my 11 year old why the question "Will robots ever have feelings" is part of philosophy, not science?

Question

My 11 year old is tasked with interpreting a Seneca quote. I started out by trying to explain to him what ethics is by contrasting 'Will robots ever have feelings?' as a question in Philosophy of Mind, with 'Should robots have rights?' a question in Ethics.

He then completely stumped me with the following back and forth:

"Isn't 'Will robots ever have feelings?' a scientific question? That's what you do for work, and you're a computer scientist, not a philosopher." (I work in Machine Learning)

"Science is about things you can observe and measure through experiments, and real feelings (as opposed to simulated feelings) cannot be observed or measured because they are locked deep inside someone's mind."

"That doesn't make sense. Science is full of examples where people asked questions about things they couldn't observe or measure, like the speed of light. They didn't just ignore them and start studying other stuff, they worked hard to come up with tricks to observe and measure them. "

At first I thought that I had just chosen a bad example: AI and consciousness is too much of an edge case, an example that is more inline with what Daniel Dennett meant when he said that scientific disciplines that are in their early stages (or as Kuhn would say, before a paradigm was established) necessarily involved metaphysical considerations.

But the more I thought about it, the more I felt that his comment was a Quine-level deconstruction of the entire empiricist mindset, not just a quirk of the borderline topic of A.I. and consciousness.

So my main question is:

How can someone defend empiricism and the scientific method against the argument that unverifiable or unfalsifiable statements cannot be dismissed as unscientific, because for all we know, we simply haven't developed the techniques or technology to observe them?

Answer 1

It is a scientific question. And a philosophical question. And those answer differently.

AI are increasingly built with 'mindsets' or emotions, like doubt, as a way to avoid ruthless 'paperclip maximiser' behaviour, like interpreting a request for a coffee as over-riding a colleagues life who is in the way of the coffee machine.

Jonathan Haidt has linked aspects of human mindsets to their cultural demands. Herding cultures like cowboys, or Afghan hill people, to honour cultures that aim to deter a raid that could take all accumulated wealth. Vs more collectivist ethics of say rice farming societies, where almost everyone had to join in planting and harvesting. He also finds that relative placing within a group on personality dimensions like tolerance/intolerance of ambiguity, open/closed to new experiences, & whether you value sanctity/degradation, can all be linked to how under threat a person has felt they have been personally, their community, or nation (on average). So we can see how mindsets respond to circumstances, and tend towards different emotion reactions and routines.

At a more basic level, we can see emotions as 'bookmarks' for physiological states, and relevant memories that go with them.

On the philosophical track, there is the Chinese Room argument. Can we separate understanding, from following rules? Where is the understanding, in the rules, applying them, updating them?

Then qualia, are they the source of the 'something extra' over a rule-follower, the aspects of subjective experience that haven't or perhaps can't be communicated.

How much can we really separate animal, human, AGI? The future, with Neuralink and genetic engineering, looks like they will fuse. Peter Singer argues moral progress is about expanding the circle of our concern. But of course that has to depend on maturity, response to freedoms.

It is a substantial question whether digitised humans will be the same kind of creature, separated from our bodies. Embodied Cognition looks at how our motor systems, & hormones, &c may be shaping our cognition.

"Science is full of examples where people asked questions about things they couldn't observe or measure, like the speed of light. They didn't just ignore them and start studying other stuff, they worked hard to come up with tricks to observe and measure them. "

Absolutely. But, they rationed their resources, according to how likely they were to pay off. Consider how the research programme of superstring theory (or m-theory) is increasingly widely seen as a failure, because it has so many variables and is so flexible it can adapt to any cosmological result, while not making unique predictions. That was decades of leeway, because it held out a lot of promises. But then alternative approaches come along, and they get the funding.

We can use the functionality of programming AIs with 'curiosity', to understand how we explore the space of new theorems in an efficient way. We need a mix of strategies, and types of investment.

Newton & others expected the speed of light 'corpuscles' to vary. Maxwell's work on uniting electricity and magnetism predicting a single speed for all light in a vacuum, and experiments confirmed. Einstein took the lack of evidence of a 'luminiferous ether', but Lorentz contraction was a serious alternative, until accounting for the orbit of mercury & predicting light would bend during an eclipse.

Answer 2

Science, given a rigorous definition of feelings, and of a robot, can tell us whether robots will ever have feelings. Philosophy is what we use to figure out the definitions of things that don't have a good enough definition yet. We don't have a good enough definition of feelings yet, so science can't be used to answer questions about it. Science can't tell us whether robots, humans, plants, or rocks have feelings, until we have a good enough definition of feelings.

Answer 3

Explaining to my 11 year old why the question “Will robots ever have feelings” is part of philosophy, not science?

Philosophy answers the most fundamental questions about how ideas fit together to derive meaning. Science usually works on the basis of existing ideas to derive deeper insights or incremental improvements to these ideas.

There seems to be two parts to this question:
(1) Will robots ever be alive (have an actual human mind)? (philosophy)
(2) Will human emotions ever be able to be simulated accurately? (computer science)

John Searle's Chinese room seems to answer the first question.
The Chinese Room Argument https://www.youtube.com/watch?v=18SXA-G2peY
Even if a machine becomes all knowing (and can perfectly simulate any aspect of a living human mind) it is still just gears and pulley's on the inside, thus not alive.

The answer to the second question would come from artificial intelligence research and seem to be YES. I have been working understanding the architecture of natural language understanding systems such as the Cyc project for a long time. Adding emotions on top of the comprehension of natural language seems quite plausible.

Answer 4

"How can someone defend empiricism and the scientific method against the argument that unverifiable or unfalsifiable statements cannot be dismissed as unscientific, because for all we know, we simply haven't developed the techniques or technology to observe them?"

Empiricical and scientific methods that are unverifiable or unfalsifiable do not meet the necessary burden of proof that they do exist. ...David Hume: "A wise man proportions his belief to the evidence"

as in the example:

We currently don't have good evidence that robots have feelings. We also don't have any evidence that cardboard boxes have feelings. So should we hypothesize that cardboard boxes could have feelings? No, because we have no evidence... But should we hypothesize that robots could have feelings? Yes, we could, because we have evidence that robots respond to specific conversation (like twitter bots). ..Notice we use empirical evidence (even what some might consider weak empirical evidence) to step into a grey area of what we can hypothesize and that science can study. ...Their is no certainty that a cardboard box could have feelings, But until their is evidence, it is not worth studying. ..The scientific method doesn't exclude ideas, rather it includes the ideas that have empirical evidence.

Answer 5

Firstly avoid speaking of science and philosophy, and find some more understandable words, "if it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck."

Then, explain that philosophy could have more discussions in such situations, and what they are. The above quote could also result from a philosophy discussion, after all.

Actually, it is part of science. But science has only one answer. It takes most sensical philosophical answers not conflicting with reality identical to the one answer, so it's not much of concern at most of the times. But in philosophy one could explore the more details in it.

There is a saying: "In theory, theory and practice are the same. In practice, they are not." It might be similar to your situation. You are not supposed to be able to accurately describe such philosophical topics in the framework of science. You could only just start right out from philosophy. In philosophy, something "is part of philosophy, not X" may not make much sense, because a lot of things started as a part of philosophy and split to a standalone domain. It's difficult to say something is not philosophy, so the fact that something is part of philosophy hardly imply anything. Or, if you don't actually care about philosophy, just say something is unscientific, not getting the word "philosophy" involved in it.

Answer 6

Unfalsifiable claims can be dismissed as unscientific if they are unfalsifiable by definition. Broadly speaking, science is, by definition, about verifying and falsifying claims about the workings of the universe. Thus, unfalsifiable claims that are so by definition find themselves outside the domain of science.

String Theory, for instance, is at least potentially falsifiable, even if it can't be falsified given the current state of technology. On the other hand, the claim that invisible, unobservable green aliens surround us every moment of our lives is by definition unfalsifiable and unverifiable since they are literally impossible to observe. Disputes about these aliens are thus not considered scientific.

Your question comes down to whether or not mental states (or felt states, or "feelings") are externally identifiable; if they are not, then one might say they are by definition externally unverifiable, and thus the assumption that "science will eventually explain consciousness" becomes null and void.

Obviously, our own conscious minds are verifiable and analysable introspectively, as Descartes made famous. We have no problem analyzing our own consciousness privately and internally.

But this is where paths split. Some would argue (e.g. David Chalmers) that consciousness is categorically (and ontologically) different from physical matter in that it is subjective, private, qualitative, etc. while physical matter is objectively, externally and quantitatively analyzable—meaning there is an unbridgeable explanatory gap between the two. And if felt states can't be identified objectively or somehow causally reduced to physical properties, then they are ipso facto externally unverifiable.

Type physicalists, functionalists, and eliminative materialists, on the other hand, simply identify mental states with physical states, and claim that they are nothing over and above the physical. If this basic view were to be correct, then answering the question "Will robots ever have feelings?" is as simple as identifying a particular set of physical conditions which are supposedly identical to feelings.

This is the core issue of your discussion. Consciousness must at least have the potential to be understood and studied scientifically (objectively, externally, quantitatively) in order to say that our current technical abilities preclude us from understanding it, but not all philosophers agree that this is the case.

Answer 7

The robots can learn/unlearn and think/analyse on the basis of electronically stored digital data(0s and 1s). The robots, however, can outperform humans in handling large data (analysis).

Nevertheless, humans are by birth innovative in perceiving/sensing, unlike robots which are limited to programming based analysis/logic. The robots can have brains equivalence storage but no human mind and intelligence. This human intelligence is based on self wish and desire which robots always lack.

By the way, there was an Indian movie released in which the robot (named chitti) had some feelings, but that was a total sci-fi, trying to dipict some sort of reinforcement machine learning.

We humans (and some other higher species) have the intelligence which has a connect to the consciousness which the robots could never be programed for. The human consciousness is beyond the five prime sense organs and perceptions.

The robots, however, can have some sort of so-called feelings (of artificial intelligence, etc) which can be totally on programming based learning/analysis with recognition and authentication by the different sensors and their actuation circuits.

Actually, we humans have a eternal desire to enjoy (like God). In this process we have designed a super faithful scientific slave (device) known as robot. And now we are trying to make robots exactly like the humans. So, there we try to set a trade off between scientific and philosophical aspect of Robotics.