Side-Effect Effect Without Side Effects and Side Effects without Side-Effect Effect

Together with Hichem Naar, we just finished writing the full version of our paper about the Knobe Effect, whose previous shorter version could be read here. The paper is about the range of the Knobe Effect and contains two experiments.

Experiment 1 refutes the idea that Knobe Effect is limited to side effects and shows that means are also in the range of the Knobe Effect.

Experiment 2 shows that good side effects in which the agent is considered to really care about the side effect are judged strongly intentional, thus suppressing the asymmetry.

You can download the file here: Download KnobeMEANS2

I'm planning to submit the paper by the end of the week. Comments are welcome.

Comments

[Joshua Knobe]

These are very surprising results, and I think they point toward an importantly different way of understanding these phenomena.

It seems like the early work on intentional action in experimental philosophy inherited the idea (from more traditional analytic philosophy) that we should be looking for some kind of 'analysis' involving something like necessary and sufficient conditions. So the idea was that one would come up with an analysis of intentional action that involved the usual sorts of mental states -- belief, desire, etc. -- but then also included something about the moral significance of the act.

But more and more, it's looking like that strategy just isn't an appropriate one here. It seems instead that moral considerations have a really pervasive effect on all sorts of aspects of cognition. This impact isn't limited to intentional action, and -- as you guys show here -- it isn't limited to cases of side-effects either.

So I am thinking that maybe we should just abandon the idea of explaining these sorts of effects by coming up with anything that looks at all like an 'analysis' of some particular concept. Instead, it seems like we'd be better off looking for a very general theory that characterizes the impact of moral considerations on human cognition as a whole. Then the goal would be to show that the individual effects observed for particular concepts (like the one you discover here) can be derived from that general theory.

What do you think?

[Florian Cova]

I agree that " we should just abandon the idea of explaining these sorts of effects by coming up with anything that looks at all like an 'analysis' of some particular concept." As expressed in the last section of the paper, I have doubts about "concept-based" explanations. Nevertheless, I don't think this necessarily means that "we'd be better off looking for a very general theory that characterizes the impact of moral considerations on human cognition as a whole." (which is a very interesting possibility). On some (pessimistic) days, I tend to think that there is a third way: maybe most of the findings about "intentional action", "causality" or "constraint" could be explained by a general theory that characterizes the impact of moral considerations on human LANGUAGE as a whole. Even if "pragmatic explanations" of the Knobe Effect (according to which people make mistakes) seem weak, there is still possible accounts according to which our words would be loaded with moral considerations. I think we still need clever experiments to determine if the "pervasive impact of moral judgment" is an impact on language or an impact on concepts.

Here is an example of experiment I would like to run after the summer. We know that, when faced with the Harm Case, people tend to say more that the Chairman's intention (or desire) was to harm the environment than in the Help Case. Do they just use a morally-loaded word ? or do these replies show that moral judgments impact on their folk psychology and their ascription of mental states ?

To determine this, we could run the following experiment. First, we give the Harm Case and ask participants if harming the environment was the Chairman's intention (or if the Chairman intended to harm the environment, or if he desired to harm the environment). Then we propose a second case in which we ask the participant to imagine that, in fact, the vice-president comes up one day later with another plan, which is exactly the same, except that it won't harm the environment. We then ask participants which plan the Chairman is going to choose.

Then we make the same thing with the Help Case.

Now, participants are more prone to say that the chairman intentionally harmed the environment. If this reflects a impact of moral judgement on their ascription of mental states, then we should observe the following thing : the tendency to predict that the "Bad Chairman" will choose the "harming plan" will be greater than the tendency to predict that the "Helping Chairman" will choose the "helping plan".

There is two solutions:

1) People says that the Harming Chairman desires to harm the environment more than the Helping Chairman desires to help the environment and have a greater tendency to predict that the Harming Chairman choose the Harming Plan than to predict that the Helping Chairman choose the Helping Plan. So, verbal responses and psychological predictions are congruent and this shows that people's use of words really reflects the way the use their mental concepts.

2) We obtain the reversed pattern: people in fact have have a greater tendency to predict that the Helping Chairman choose the Helping Plan than to predict that the Harming Chairman choose the Harming Plan. So, "mental words" are in these cases disconnected from "mental concepts".

I must admit that I don't know what the results will be like.

[Jonathan Webber]

About your proposed experiment, in which the Chairman is offered an alternative plan:

You say that if the Knobe Effect reflects the impact of moral judgement on the ascription of mental states, then we should find that participants are more likely to predict that the Bad Chairman will choose the Harming Plan than to predict that the Helping Chairman will choose the Helping Plan.

But surely this is only the case if the mental state in question is the desire to harm the environment. What if negative moral judgement drives the ascription of a belief, such as the belief that harming the environment is unimportant?

This should give you the third experimental outcome that you exclude in your posting: that people are equally likely to predict that the Bad Chairman changes plan as to predict that the Helping Chairman does. (Though this outcome would be compatible with other accounts of the Knobe Effect too.)