本文探讨了作者对责任概念的深刻理解，从最初的肤浅认知到如今更为成熟的看法。文章指出责任并非客观存在，而是人类基于共同理解而赋予的工具。作者通过亲身经历，分析了责任如何被用作逃避行动的借口，并反思了“承担一切责任”的弊端。最终，作者提出了一套实用的责任观：在明确他人责任的前提下，不以责任为借口而无所作为，并在必要时承担“备用责任”，以促成更高效的协作和更积极的行动。这种观点平衡了个人的责任感与实际行动，旨在避免责任的滥用和推诿。

🤔 责任是人为构建：责任并非客观存在，而是人类基于特定规则和共识赋予的工具。我们根据实际情况将责任分配给他人，这是一种社会契约，并非自然规律。

🙅 责任不应是借口：我们常常以“不属于我的责任”为借口逃避行动。然而，当他人未能履行责任或无人负责时，我们应该积极介入，而不是袖手旁观。例如，对野生动物的苦难视而不见，是一种不负责任的行为。

🦸 全盘负责的误区：虽然我们不应逃避责任，但承担一切责任也会带来负面影响，如难以放松、导致效率低下，甚至可能演变为控制欲。因此，合理的责任分配至关重要。

🤝 实践中的责任：在实践中，我们应该适当地将责任分配给他人，特别是当事情不涉及灾难性后果时；同时，当出现问题时，不应以责任为借口而无所作为。在必要时，承担“备用责任”以支持他人。明确责任有助于提高效率和协作。

🧐 责任的开放性问题：尽管提出了实用的责任观，但仍有一些开放性问题需要进一步研究，例如如何在团队决策中平衡个人意见和集体智慧，以及如何避免因责任分配不均而导致的决策失误。

Published on January 21, 2025 10:47 AM GMT

My view on the concept of responsibility has shifted a lot over the years. I’ve had three insights that brought me from my initial, very superficial and implicit understanding of responsibility, to the one I have today, which I consider more accurate, more practical, and more healthy.

 

 

Responsibility is Made Up

The first insight came while I was part of a dinner debate with a local philosopher and a few friends. We were discussing AI, of course, and the philosopher eventually brought up the question of responsibility: if a self-driving car malfunctions and ends up killing a pedestrian, who’s responsible? The human in the car? The car manufacturer? The programmers? The car itself? Can an AI ever be “responsible” for anything?

Considering this, I concluded that the question is a bit misguided. It sounds like a question about the world – as if the responsibility just lies somewhere and we can figure out where if we look hard enough. But at the end of the day, responsibility is just a tool that humans assign to others based on certain heuristics and shared understandings^[1]. This doesn’t mean responsibility is not a practical tool. You can define it in certain ways and apply it in society, and if you do that in the right way then it ends up beneficial. But it was still a valuable insight for me that responsibility does not “exist” but is just a more-or-less-agreed-upon construct. And hence “Can an AI ever be responsible” is not so much a factual question. It’s not a question about the AI. It’s a question about when it becomes practical for humans to assign responsibility to an AI.

Responsibility as an Excuse

Let’s look at two scenarios that I encountered:

We were planning with around ten friends to meet the next day to get picked up by a bus. The bus would arrive at 10:05 the next morning, so Arnold suggested we meet at 10:00 at the bus stop. At which point Bianca said “Surely somebody will be too late then, let’s rather say we meet at 9:45 to be sure we all make it?”, to which Arnold responded “But then that’s the late person’s responsibility! Everybody should themselves be responsible to be on time”. I get where Arnold is coming from. But still, you’re deciding between two outcomes: We choose 10:00, which probably means somebody will miss the bus and hence miss out on the trip and everyone will be very sad about that and that person is responsible for that. Or, we choose 9:45, which probably means everyone will make it to the bus. Scenario 1 is most likely worse (unless you have significant opportunity cost – but then you could still unilaterally decide to just come at 10:00. 9:45 would just be the “Schelling time” after all, and not some unbreakable commitment), and the fact that you know who is to blame for it doesn’t make it any better.I once discussed wild animal suffering with someone, let’s call her Claire. I made the basic argument along the lines of: there’s an unfathomable number of wild animals out there, and the majority of them get eaten or starve or die in other horrific ways, often just shortly after being born. If there’s even a relatively low chance of them being able to suffer, this is a huge deal. This is also an extremely neglected problem that only a handful of people worldwide are looking into. I don’t know if it’s easy to make progress there, but it seems like humanity should take the problem much more seriously than it currently does, because if there are ways to improve this situation without major drawbacks, then it would be quite a tragedy not to act on them. Claire was skeptical and argued that this is not humanity’s responsibility – we should rather solve factory farming and human suffering, as these are problems we’re actually responsible for. But wild animals would exist anyway and are not very affected by us, so we should leave them to themselves. To which again I argued that there are two scenarios: Either we leave wild animals to themselves, which means they keep on suffering just as much as they’ve always done. Or we figure out some ways to alleviate that suffering and make the world just a bit better in the process. The animals don’t care who’s “responsible”, they just suffer. If I see an injured man lying on the side of the street, I don’t just walk past because I’m “not responsible” (well, at least I have the ideal of not doing that, and certainly hope that I would live up to it if I ever were in that situation). I would do my best to help him out, because I happen to be in a situation where I can help him, and he needs help. And to me, humanity and wild animals are kind of like that. And hiding behind our apparent lack of responsibility to me just seems like a convenient excuse to do nothing about the problem.

After this realization, I turned slightly cynical, considering responsibility a flawed concept that just leads to a lot of bystander effects and finger-pointing. Knowing that somebody else is technically responsible for something conveniently also means that you don’t have to get your hands dirty.

One way to react to these insights would be to become nihilistic about it, disregard responsibility altogether, and hence feel responsible for nothing. Another way to react is to feel responsible for more or less everything. Considering myself a positive and altruistic person, I went with the second option. And I suspect this conclusion is not all that uncommon in effective altruism: many of us are pretty consequentialist, thinking mostly about which problems exist and what actions we can take to solve them. The preferred decision-making tool is usually not responsibility, but counterfactual reasoning. We think less about whether we should feel responsible for these problems or just leave them to someone else. Rather, if our actions lead to better outcomes overall than our inaction would (and we have nothing even better to do), then that’s reason enough to act. One might call this heroic responsibility.

And this is where I then spent a few years, thinking responsibility as humans typically use it was a flawed concept, and hence feeling responsible for everything. So I tried following all the promising directions in my action space that plausibly improved things for the better. But, as it eventually turned out, that’s not optimal either.

Feeling Responsible for Everything is a Slippery Slope

While the consequentialist in me was happy with my renewed understanding of responsibility, I couldn’t help but realize that it had some drawbacks as well:

Feeling responsible for everything makes it difficult to ever relaxResponsibility, when done well, is actually a good way to coordinate; if many people feel responsible for everything, that can introduce inefficient redundancies and people stepping on each other’s toesFeeling responsible for the actions of other people can be patronizing and take away their agencyEven worse, it can eventually turn into manipulative behavior^[2]

And these are some serious drawbacks! So, after these three insights – responsibility being a made-up human construct, one that is occasionally used as an excuse to do nothing, but also one that is useful nonetheless, because claiming all the responsibility yourself has some serious negative side effects – I ended up where I am today.

Practical Responsibility

All things considered, I nowadays try to stick to the following heuristics:

Leave some responsibility to others. Don’t take everything yourself. Particularly if a thing is unlikely to have catastrophic consequences, it can be a good trade-off to leave responsibility for some particular thing fully to others, assuming they want it and it’s clear to everyone involved that they are the ones bearing it. Even if things end up going badly, this can have learning effects that people would otherwise miss out on.Don’t let responsibility be an excuse to do nothing. When there’s a problem that either nobody feels responsible for, or the responsible party is clearly not living up to their role at the expense of others, then it may be worth intervening. Whether on a global scale, such as with wild animal suffering, or a local scale, such as an injured person on the side of the street that others are ignoring.Point out who’s responsible when necessary. A common failure mode is that a group of people settles on a decision that requires some action, but they don’t decide who’s ultimately responsible for overseeing that change. Once it’s common knowledge who’s responsible for something, this both increases the likelihood of that thing happening and gives the other people peace of mind, as they otherwise might feel just responsible enough to get distracted by or feel guilty about it, but not necessarily responsible enough to actually see it through.When easily possible, feel free to take “backup responsibility”. One example I come across often is bringing a spare HDMI cable when meeting for a talk in a seminar room. Relatively often these are missing or broken, and even when I’m not in charge of running that event, the cost of putting my HDMI cable into my backpack is almost 0. So even though I’m not officially responsible, I can have the responsible person’s back here very easily without infringing on their agency. I’ve also experienced this from the other side, where I was hosting an event but forgot to bring a cable myself, at which point I was grateful that two other attendees had felt responsible enough to bring one just in case.

Admittedly this is not an all-encompassing solution. Some frictions remain. E.g. I still find it occasionally difficult to leave important decisions to a group of people, when I don’t fully trust in these people’s judgment. I tend to struggle a lot with then figuring out what’s less bad: deliberately influencing the decision, e.g. via means of nudging or framing, or accepting that we likely end up making a suboptimal decision?

So some open questions certainly remain. “Further research is needed”. But accepting responsibility as a made-up yet useful tool, that should be shared with others, and that should not be abused for justifying inaction towards problems, seems like a much better approach than the many points along the way that I traverse on my way here.

What’s your relationship with responsibility? Are you the heroic, the nihilistic, the pragmatic type, or something else entirely? I’d be interested to hear other people’s views, and in case you, like me a few years ago, haven’t thought much about the topic, maybe consider this an opportunity to do so.

---

Thanks to Adrian Spierling for his insightful feedback on this post.

1. ^{^}

   I’ve been told that some would disagree with this view. I hereby mention that this is the case and that my views are of course disputable. I would still assume that, for many people who haven’t reflected much about the concept of responsibility it would resolve some confusion to start considering it a made-up thing that humans use for practical reasons, instead of something that objectively exists in the world. Even if, similar and related to moral realism, one can certainly make the point that there really is some objective thing that maps to what we call "responsibility".

2. ^{^}

   To clarify, I’m not talking about galaxy-brained puppet master manipulation tactics here (although eventually, one might get there as well). In line with the idea of social dark matter, manipulation of others is probably much more widespread as well as usually much less severe than many assume when they think of “manipulation”. So feeling responsible for everything does not necessarily lead to being all plotting and deceptive, but it certainly may have the effect of, much more than one would otherwise, considering how one's words and actions will impact how others react, and then communicating and acting in ways to make it more likely that others react in preferred ways. I think there are versions of this that are very widespread and entirely harmless, even beneficial (e.g. being very careful and deliberate when giving negative feedback, to increase the likelihood of the other person taking it well), but the more space this takes in one's mind, the easier one drifts into much more questionable waters – a slippery slope.

Discuss