Side Effects Revisited

In an earlier paper I tentatively floated an account of how ascriptions of intentional action might work in side effect cases.  If I modify it slightly, I get the following:

In cases involving side effects, an agent will (normally) be judged to have intentionally brought about a side effect, y, by performing some action, x, only if the following conditions are met:

(1) The agent (a) wants to do x, (b) wants to bring about y by doing x, or (c) both (a) and (b).
(2) The agent knows (or at least believes) that doing x will bring about y.
(3) The agent intentionally (a) does x , and (b) brings about y, roughly as planned.
(4) If y is bad, the mental states (i.e. desires, beliefs, intentions, etc.) of the agent must be such that the agent is perceived to be blameworthy.
(5) If y is good, the mental states (i.e. desires, beliefs, intentions, etc.) of the agent must be such that the agent is perceived to be praiseworthy.

As far as I can tell, doesn't this settle with most (if not all) of the existing data so far?  What am I missing?  Thoughts anyone (or everyone)? 

 

Comments

[John]

Hi Thomas,

About the normality condition. How do you want this to be understood? Oftentimes it just means the same thing as 'typically', which seems to be a sort of statistical notion. But I gather that in, e.g., the generics literature, 'normally' is construed as a modal notion. It strikes me that there might also be other ways of construing it -- e.g., epistemically. Did you have any particular construal in mind?

Also, about (4) and (5): does y actually need to be good/bad, or is it enough for your purposes that it's perceived to be good/bad?

[tnadelhoffer]

John,

Thanks! Yeah, (4) and (5) should have said, "If y is perceived to be good (or bad), the mental states of the agent must be such that the agent is perceived to be good (or bad)."

All I was really aiming at with the "normalcy" bit was that you won't ever get a list of necessary and jointly sufficient conditions when you're trying to capture how people use a concept such an intentional action. At best, you'll hit on something that explains how most people can be expected to answer given certain conditions. As such, I suppose I meant it more like "typically" as you suggested.

[Clayton]

When you say perceived as good (or bad), I want to know perceived as good or bad by whom?

It seems that in some of the stories that we use to elicit the intuitions underlying the side-effect effect, the effects are bad to the agent, sometimes they are bad to the evaluator, but there doesn't seem to be some fixed standard for determining whether they are bad. At least, that's what I've found. So, it should be something like 'bad relative to some contextually determined standard'.

[tnadelhoffer]

Clayton,

When I floated this analysis a few years ago, I was only interested in folk ascriptions of intentional action in side effect cases. As a result, I was only interested in how participants in the studies perceived the side effect, morally speaking. I nevertheless think it would be helpful to ask participants two questions--one about whether they find the action (or side effect) to be bad, good, or neutral and one about whether they think the agent finds it to be bad, good, or neutral.

I have suggested that the difference between these two questions might explain why you get the asymmetry in the HARM vs. the HELP cases. I think that in the HELP case, participants judge that helping the environment is good, yet they judge that the chairman does not think it is good--after all, if he thought it was good, why wouldn't he care that implemening the program would help the environment? Because they don't think he cares about helping the environment, they don't judge that he helped the environment intentionally. Indeed, in some sense, the chairman in the HELP case seems even less moral than the chairman in the HARM case.

One way to see whether I am right about this would be to ask participants the two aforementioned distinct questions.

[Clayton]

I've had some success getting students to say that unintended side-effects are brought about intentionally even though the subject isn't to be blamed for having acted in the way that she did using two types of survey.

First, I offered them a case of moral conflict in which the students said that the effect that the agent knew she'd produce but didn't intend to produce was brought about intentionally even though they say that they woudlnt' blame the agent because she did the right thing.

Second, I offered them a case of no moral significant whatsoever in which someone is purchasing a car and after being told the price is told that the car is a color the buyer concedes she doesn't like. The students again said that she intentionally purchased, say, a car that was yellow with no intention to purchase a yellow car even though blame does not factor into it.

The surveys I've run suggest that there is some empirical evidence against (4) (as does Knobe's survey on the Nazi's racial identification laws).

Also, I'm curious about (3). You seem to be suggesting that side-effects are brought about intentionally only if they are brought about (roughly) as planned. I thought that the standard view among those who analyze intentional action in terms of action in accordance with a plan (e.g., Brand), the plan for action includes provisions for bringing about an end, but it is no part of the plan that a side-effect is brought about (Brand is explicit that foreseen is insufficient for planned). Might it be better to revise (3) so it says "roughly as foreseen" rather than "roughly as planned"?