Anthropologists tell us that home sapiens are concerned about "others," even after they die.

In settled (as opposed to nomadic) societies this concern seems to catalyze beliefs, rituals, and institutions, taking many forms.

This is so even though the ancestral dead have no further empirical or "use value."

To restate, the ancestral dead are, obviously, the origin of all knowledge, concepts, and values to the living. They transferred all the information. But the "value" of the "source of information" is abruptly terminated once they are dead.

Is this "care for the ancestral dead" an emotional or a rational reaction? If AI captures human consciousness, would it have to model "care for the dead"? How could it rationally go about this?

  • 2
    I think this question is too big for SE as written. While I've got some sympathies for the answer. How would we determine what a good answer is to this question? It ranges over many things and asks both why we as people value the dead and whether an AI must do so to capture human consciousness.
    – virmaior
    Nov 13, 2015 at 5:09
  • I must leave it for now, but I do agree with your comment. I am trying focus a question about the dead that avoids the usual cultural-anthropological answers and places it more in terms of the circulation of "reason." Part of my interest is in the problem of how one defines AI and how one might generate "values" in a computer. Concern for the dead also avoids the usual utility rationalizations of altruism or love. Anyway, I need to rethink it, but will leave for now. Nov 13, 2015 at 13:52
  • Also just to note, from the AI-view this forms an interesting paradox of "essential information that also cannot be recovered." A value for "loss of information" that is somehow defining of consciousness. All its logic would also need to be organized at a secondary level with some sort of "echo effect." But now I'm just thinking to loud, sorry. Will revise when I can. Nov 13, 2015 at 14:20
  • 1
    What is the basis of your implication that nomadic people do not do this?
    – Dave
    Nov 13, 2015 at 14:45
  • In "Structure of World History" Kojin Karatani discusses the anthropology of this, though I do not have his references to hand. This cultural-historical shift in attitudes towards the dead (both concern and fear) was partly what prompted my curiosity. In reality, I am sure this is quite controversial, and obviously access to primary "uncontaminated" nomadic culture is bound to be very limited. Though Neanderthals and some animals show "concern for the dead" it is not obviously universal to "rational beings." Nov 13, 2015 at 15:05

3 Answers 3


People don't magically disappear from our imaginations when they die. If we valued their thoughts, beliefs, attitudes while they were alive (and even when not immediately present), we will still tend to do so even when they are deceased. We don't need them literally present for them to re-iterate their wisdom; we can remember it, or imagine what they would have said or done. As with most important mental phenomenon, this is a mix of rational and emotional responses -- the continuing feeling of attachment is what keeps our memories of the deceased accessible and salient to our lives.

From this view of the phenomenon, we can extract a few things that an AI would need in order to mimic this type of effect: As a minimum requirement, it would need to be able to remember what the now deceased agent (it might be another AI!) did, even better would be a model of what the agent would do. The AI could then incorporate that information into its own cognitive (-ish? for people who deny strong AI) processing. If it has introspective capabilities, it might rate that keeping this information around is valuable (with respect to some goals/measures).

Aside, even if the AI could upload the full state of some other agent, a similar effect could still arise due to resource constraints. If the AI is up against its resource constraints, it may not have the ability to have all of the it's "ancestors" knowledge fully loaded. A sensible thing to do would be to index the reasons why the AI would call on a particular ancestor (and use that index to decide when to load its "ancestors"); that index would comprise the analog of the salient memories.

  • One thing to consider: its often hard to tell if something is actually dead or not. In many cases, having a one-click purge of the memory of something because it "died" could be harmful, wasting valuable information that might have been saved by a more gradual process.
    – Cort Ammon
    Nov 13, 2015 at 6:13
  • A real life example of such AI algorithms can be found in tracking algorithms. If an object disappears from view, many algorithms will keep the memory of the object alive as a "persistent track" for a while. Such a persistent track is used later to help make sense of additional detections, particularly in the case where an object had simply gotten hard to see (such as it went behind another object) but has now returned to view.
    – Cort Ammon
    Nov 13, 2015 at 6:21

It seems to me very easy to explain care for the dead as an evolutionary spandrel, i.e. an unintended side effect of naturel selection:

  • Care for our living relatives and elders was a useful adaptation for the purposes of collaboration and transmission of knowledge and skills from generation to generation.
  • Belief in the afterlife comes from religion as a defense mechanism/religion as Freudian neurosis theory: people had to believe in an imaginary afterlife as the only way to cope with the pain and randomness of violence in the real world.

The combination of these two adaptations then leads very naturally to a care for the ancestral dead, if only as means of reassuring oneself that belief in the afterlife is indeed justified.

As for your second question regarding AIs having to care for their "departed elders", mostly likely such a belief wouldn't be necessary for an AI. Presumably by the time we are proficient enough in how the mind works that we are able to program sentient AIs, we will also know enough that we can avoid the various quirks and neurosis that plague human psyches, and AIs will be free of any of the unnecessary religious baggage that humans have to carry around.

There is one scenario where belief in an afterlife and care for the dead ancestors might be necessary for an AI. If it turns out that a survival instinct is a necessary condition for sentience. For example, based on Daniel Dennett's intentional stance, it might turn out that autonomous goal seeking behavior can only be implemented by programming a "will to live" into these AIs, and somehow this survival instinct inevitably leads to a belief in an afterlife and a care for dead ancestors. Of course for AIs, belief in an afterlife (or more precisely reincarnation) seems much more justified than it does for humans. All it takes is for some one to load a dead AI's backup tapes and voila! reincarnation!!!!

  • Yes, I picked "concern for the dead" because it seems unusually lacking in rational utility, unlike many instincts or emotions. It appears to affect some animals, while being of far less concern to nomadic humans. I like your answer, except all the interesting "devilish" details are missing, and we all tend to cover up too much by appeal to Freud. You are right to point out the AI would not "die." Thus it might have no concept of "essential information that can't be recovered." It would need some sort of "echo effect" attached to all information. Is this a mind-defining attribute? Nov 13, 2015 at 14:12
  • @NelsonAlexander "devilish" details? please enlighten me. Nov 13, 2015 at 17:43
  • Probably not worth going into here. But, for example, "...we can avoid the various quirks and neurosis that plague human psyches, and AIs will be free of any of the unnecessary religious baggage that humans have to carry around." Well, not so fast. Isn't AI supposed to mimic those? And "freudian defense mechanism" and "evolutionary adaptations" seem to me like dogmas covering up the real steps. Nov 14, 2015 at 2:32
  • "Isn't AI supposed to mimic those?" No - there's no a priori reason why it has to. Computers where originally meant to simulate humans performing calculations and information retrieval, both of which they perform without the errors and issues that humans face when they perform those tasks. There's no reason why the same won't apply to full fledged AI as well. Nov 14, 2015 at 6:45

There's both an emotional and rational component; for Simone Weil, in her book On the Need for Roots (L'Enracinement) the future - as then too, the present - is constructed out of the past; and all the 'treasures of past' - primarily, language and ways of living - and our inheritance - are neccessarily created by the ancestral dead over an epochal range.

  • 1
    I guess I am trying to separate out the rational and emotional here, and then rationally separate the "dead ancestor" from, say, a "lost book." Need to rethink the question. Nov 13, 2015 at 14:08
  • 1
    @Nelson Alexander: I suppose there is a difference between ones personal ancestors, and cultural ancestors - which might have some consonance with Jungs Archetypes and collective unconscious; interestingly, at least for me, given that I was educated as a mathematician, there is a mathematical genealogy project! Nov 13, 2015 at 23:18

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .