Some formulations of humanism already grant moral consideration to sentient non-human animals (e.g. https://humanists.international/what-is-humanism/). Does humanism also extend to granting rights to AGIs, should they become sentient?

Sentientism is a closely related philosophy that makes this explicit https://en.wikipedia.org/wiki/Sentientism.

  • $\begingroup$ We do not prmote any product here. Your question is deemed as spam. $\endgroup$
    – user9947
    Mar 7 '19 at 9:42
  • $\begingroup$ There is no product here. It's a genuine question. The FB group is just a discussion forum. I'm new to SE so apologies if I've framed this wrong. $\endgroup$ Mar 7 '19 at 12:13
  • $\begingroup$ Hmmm....The thing is that we try not to promote personal blogs, websites or codes. If we do we must declare our affiliation and even then it is up to the moderators to decide. Also in SE sites we try to frame our thoughts in a question and not an open discussion. You can tour the guidelines of SE for a better understanding of the rules. $\endgroup$
    – user9947
    Mar 7 '19 at 12:17
  • $\begingroup$ Thanks @DuttaA. I've tightened the question and removed the links to my other article + the community web site. Hope this is better. $\endgroup$ Mar 7 '19 at 16:23
  • $\begingroup$ I like this question and will answer if I have anything to add to Neil's. (My sense is there is a deeper question of whether humanity, here in the general sense of Ren , is an exclusive characteristic of homo sapiens. Evolutionary game theory suggests that it may not be. PKD explores this in a Christian context of empathy.) $\endgroup$
    – DukeZhou
    Mar 7 '19 at 20:12

Does humanism also extend to granting rights to AGIs, should they become sentient?

I think that is going to be something that will divide self-identifying humanists into a few different groups, based on understanding about what AGI is.

The front page you link includes "aiming toward the welfare and fulfillment of living things", which does not explicitly include non-living sentient things. However, the humanist philosophy naturally extends to granting rights and protections based on reasoning and scientific evidence to anything that can be argued as needing them.

The big problems are:

  • Recognising and measuring sentience.

  • Attaching moral value to events that happen to artificial entities, where subjective measures such as pain, stress, happiness, fulfilment might not be relevant.

These measurement problems already cause a lot of variability in moral stances towards farming animals for instance. These biological creatures share enough with humans that even though we don't fully understand the nature of their subjective experience, it is possible to extrapolate with some backing from scientific studies. For example, we can measure whether a creature suffers stress or pain and reacts similarly to a human, and even if we don't understand what that means subjectively (what it feels like to be such an animal), a combination of Occam's Razor and Precautionary Principle can be argued:

  • Occam's Razor the simplest interpretation is that if an animal is based on same cellular structure, has same brain regions for experiencing pain and reward, uses same stress hormones etc, that its subjective experience is likely somehow comparable to a humans. This allows us to apply empathy to an animal's situation with a guess that doing so is not 100% inaccurate.

  • Precautionary Principle the consequences for inhumanely treating a sentient creature are morally worse than the waste of humanely treating a non-sentient creature (provided doing the latter is not causing undue suffering to some other thing). So faced with a lack of precise knowledge, there is still a strong rational ethics argument for extending rights and humane treatment if there is any uncertainty.

When it comes to artificially created entities, we lose the first argument of Occam's Razor, and can only get it back with a better understanding of what sentience actually is, and which parts of it we should value. We need a much firmer theory here than with animals, because there is so little shared between our construction and that of a human-built machine. In addition, just as there are many different kinds of organism which exhibit different amounts of apparent intelligence and possible sentience, there will be different versions of AGI, probably starting with purpose-built non-narrow AIs that combine features that are practically useful and make them seem intelligent (parsing human speech and responding appropriately, which Google Home, Alexa, Siri already do to some depth) with some common-sense understanding of broad task portfolios. These things are near future, and most developers who work on them would not consider them sentient - however, they will be designed to act as if they were, because that makes them practical and usable, which in turn will make it much harder to argue Occam's Razor due to similarity in behaviour.

We still keep the second argument, but it is confounded by not understanding what the analogs are for humane and inhumane when dealing with artificial entities. It is unlikely that anyone would have a problem with switching off a running instance of the next 2 or 3 generations of Siri assistants - but at what point does that become deliberate incapacitation of a sentient being, and does it even matter if the system is designed to be switched on and off, and expresses no particular desire to be in either state?

In summary, the humanist philosophy could extend to cover the welfare of artificial general intelligences, but it is far from clear whether it can go there in practice. Until we have more complete theoretical models of AGIs, or practical working devices, then it is premature to consider it. Most debate is going to be based on imagined qualities of AGI, and any two people discussing this are likely to be starting from different assumptions about what those are.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.