Loneliness and Empathic AI
Details
Many people feel lonely in today's disconnected world. And this loneliness can create deep feelings of anguish and suffering. Does empathetic AI have a potentially positive role to play in helping some people abate such feelings? Or would turning to AI more likely make the problem even worse? Come along and arm-chair theorize with us about this fascinating topic Friday evening at Lake Ontario Park at 6pm.
Paul Bloom has a great discussion of this topic here:
https://www.youtube.com/watch?v=EpacFN8cU-Y
And he co-authored a recent paper on this issue (abstract only here), those really ambitious can download the full paper from here.
Please be sure to bring a lawn chair, we will be sitting under a tree just past the sheltered area where the bathrooms are located.
This Friday evening meetup is only for humans, so you have to leave your AI companion turned off for the duration of the discussion ha ha! Should be a fun topic, hope to see some of you there.
Cheers,
Colin
** Update.... a few reflections to help set the stage for this discussion ...
Many people have “unmet” needs that diminish their wellbeing, be those needs intellectual, romantic, sexual, physical, joy/amusement, emotional, etc. And there can be healthy and unhealthy ways of meeting these needs.
Unhealthy strategies are those that prove, in the long run, to be detrimental to a person’s wellbeing. For example, a person might try to fulfill their unmet needs for companionship by befriending someone who is exploitative (e.g. an alleged “friend” who is always asking for money). Or they may seek to satisfy their sexual desires in ways that jeopardize their health (exposing themselves to STIs). Or they overindulge in satisfying one need (e.g. amusement) at the cost of others. For example, by binge watching TV shows all day but not getting enough exercise, or texting on their phone all day to be amused but ignoring their frustrated partner sitting beside them!
Before turning to the issue of empathetic AI, we should recognize we outsource the meeting of many needs to technologies of different kinds, things other than direct “person-to-person” connection. And not all instances of this are problematic, indeed some are effective and sage strategies. For example, some intellectual needs might be met through reading non-fiction (like philosophy, written by people who died hundreds of years ago). Few would insist that one’s intellectual needs should only be met by interacting, face-to-face, with stimulating people within their proximity. But it may be a matter of degree. Someone who only tries to satisfy their needs for intellectual stimulation by reading, versus having any conversations with people in real life, is perhaps sub-optimal in their pursuit of intellectual growth, and loses out on other important needs (social connection).
Or consider one’s need for exercise. For some people, these needs might be better realized by relying on apps that measure steps or show one’s progress by comparing pace /distance on runs, etc. In such cases, we wouldn’t insist that they hire an athletic coach instead, so that they more “authentically” meet their need for physical exercise. The “proxy” of an exercise app is cheaper, and, while perhaps less effective than a coach, better than nothing.
Some needs for romance might be met through reading fiction, and some needs for sexual gratification through pornography, etc. Few would take the puritanical stance that all of your needs should be met only through direct, person-to-person, connections.
There is another important consideration. People’s access to a running coach, or intellectual interlocutor, or romantic or sexual partner, will also vary. There is no “equality of opportunity” that holds for all people in terms of being able to meet their needs. And the needs people have can also vary dramatically.
It is against this backdrop that I think one can begin to form defensible judgements about the potential harms and benefits of emphatic AI. We must first acknowledge that people’s needs for emotional connection can vary significantly. And so does their access to the human and social capital required to meet those needs by non-AI means.
Looking forward to our discussion.
Cheers,
Colin
