Monday, May 4, 2015

What you can't expect when you're expecting

Laurie Ann Paul has a recent paper on the question of whether it is possible to decide rationally whether to have a first child. She argues that it is not possible. When I heard about the paper I was very curious, because it seemed to me that it might have something useful to contribute to thinking about about the issue of choices between supposedly incommensurable alternatives.

Various philosophers and others have argued that at least some choices do exhibit incommensurability. (For a review see this entry in the Stanford Encyclopedia of Philosophy.) If they’re correct, or if at least some versions of their views are correct, then it is not generally true that all choices involve alternatives represented in a single common value scale. I don’t have a hat in the right either way, although I suspect both that there are at least some kinds of choice that defeat commensurability, and that the choices in question are of quite restricted and special types.

Although I’m not actively working on the topic of incommensurability right now, I’m keeping an eye on it. So here is a quick overview of Paul’s paper. It wasn’t actually what I was expecting at all, and after outlining her argument I’ll say a little about why. I’ll discuss the paper under Paul’s own headings. (Any heading below with a number corresponds to a section of her paper. All quotations are from Paul’s paper.)

1. Deciding Whether to Start a Family

Paul sets the scene: the decision problem we are to focus on is a person, until this point childless, who is ‘personally, financially and physically able’ to have a child. The problem, that is, which we are to consider is a kind of ‘best case scenario’ for someone thinking about whether to have a child, except for the fact that person or couple considering the question have not done it before.

And on a fairly commonly held view, what one should do is think about the consequences of either having a child or not having a child, and then choose what on balance seems the better of the two options in the particular circumstances, and for the particular possible parents. Not everybody decides in the same way.

2. Decision Theory: A Normative Model

In this section Paul introduces an informal version of decision theory considered as an account of rational decision making. Making a choice rationally involves determining (as far as we can) the outcomes that would follow each option, and then working out the expected values, for the deciding agent, of the various possible outcomes, so that we can determine which is best.

Paul claims that when choosing whether or not to have a child, one is partly choosing between options one or both of which are to a significant extent about ‘phenomenal outcomes’, or with what it would be like to have a child.

(Phenomenal outcomes aren’t the only relevant outcomes. Paul isn’t saying that, but rather that they are among the important outcomes.)

3. What Experience Teaches

Paul then endorses the view that phenomenal knowledge, that is knowledge of what something is like, ‘can (practically speaking) only be had via experience.

To bolster this view she describes the case of colour-deprived Mary, from Frank Jackson’s thought experiment. Mary supposedly knows everything there is to know about the physics of colour perception, but has been deprived of first hand colour experiences:
“As Jackson points out, when Mary leaves her cell for the first time, she has a radically new experience: she experiences redness for the first time, and from this experience, and this experience alone, she knows what it is like to see red.”
Paul’s argument isn’t supposed to depend on whether Jackson’s imagined case is telling against physicalism. Rather, we are simply supposed to recognise that Mary is in an ‘epistemically impoverished position’. (I’m not sure that even this will work, because part of what is at issue over that Mary is indeed whether knowing all the physical facts puts her in a position to know what it would be like to see red.)

Anyway, difficulties aside what Paul wants and asserts is the notion that seeing red for the first time is ‘epistemically transformative’. Not only that, when you don’t know what something is going to be like, you also don’t know how it is going to make you feel:
“This means that, when Mary chooses to leave her black-and-white cell, thus choosing to undergo an epistemically transformative experience, she faces a deep subjective unpredictability about the future. She doesn’t know, and she cannot know, the values of the relevant phenomenal outcomes of her choice.”

4. The Transformative Experience of Having a Child

Now we return to the question of choosing whether or not to have a child:
“A person who is choosing whether to become a parent, before she has a child, is in an epistemic situation just like that of black-and-white Mary before she leaves her cell. Just like Mary, she is epistemically impoverished, because she does not know what it is like to have a child of her very own.”

Having a child, Paul argues, is a distinctive and intense experience, the phenomenal properties of which cannot be known in advance. She spends most of this section fleshing out and offering some defence of this claim.

5. Choosing the Ordinary Way is not Rational

The way the argument comes together at this point should be pretty clear given how the earlier stages are set up. The standard decision theoretic model assigns expected values to outcome states. But in at least some cases, including having a first child, the phenomenal outcomes aren’t knowable in advance, and so values cannot be assigned:
“When a decision involves an outcome that is epistemically transformative for the decision-maker, she cannot rationally assign a value to the outcome until she has experienced the outcome.”


Paul also has a section (6) in which she considers a number of objections.

I said near the outset that this wasn’t really what I was expecting. That’s mostly because of how my own concerns and interests led me to interpret her title and abstract. I do think that there might well be ‘transformative experiences’ from the perspective of decision making. That is, that there might be experiences after which the decision maker is transformed qua decision maker. What I mean by this is transformed in the sense that after the experience the decision maker has different preferences, or at least a suitably substantial revision in the relative weighting of the preferences that she already had.

Having a child plausibly does this to at least some people. I think it might have done so to me, and in ways that surprised me at the time. (For example, thinking of myself as a parent seemed related to my suddenly becoming a much more conservative driver, even when I was driving alone. It just seemed more important to be careful.) If something like that does happen, then it does indeed present a kind of obstacle or challenge to some standard models of decision.

But I don’t see that the obstacle depends very much, if at all, on ignorance about what it will be like after the transformation. In fact some ways of understanding the notion of transformative experience, seem to be obstacles to standard models of decision-making even given full information. That’s because many standard decision and learning approaches focus on maximising something (goal achievement, subjective utility, reward in some reinforcement learning regime, etc.). And I’ve not — at least yet — seen one that tries to work out how to handle decisions where the options include actions that transform the maximisation target itself. (This could be ignorance on my part, but some superficial poking around in the reinforcement learning literature didn’t turn anything up. I’d welcome pointers to relevant literature in the comments below.)

Notice that it’s not obvious that all cases where options can change what a person wants will be difficult. It’s not hard, for example, to find people who would gladly sign up for an intervention that made them stop wanting to smoke but otherwise left them pretty much the same. But it does seem as though there’s something interesting and difficult about the problem of how, given a current value function, to choose between options some of which include the consequence that ones future value function, and with it the consequences of future choices, will be different. Then, even if you do know what the differences will be, it isn’t obvious what counts as making that kind of decision well, or how one might go about trying to get a robot to do it efficiently.

To be fair, Paul does mention the possibility of preference changing at one point in her paper, where she says that having a chid is not merely epistemically transformative but ‘personally transformative’ and that: “ A personally transformative experience radically changes what it is like to be you, perhaps by replacing your core preferences with very different ones.” But in this paper, at least, her main focus is on epistemically transformative experiences, focused specifically on phenomenal knowledge.

I’m hoping to give more serious attention to the topic of possibly incommensurable options in a year or so. When I do that, I’ll follow up and look at Paul’s wider account of personally transformative experiences.

Here’s the text of Paul’s abstract:

It seems natural to choose whether to have a child by reflecting on what it would be like to actually have a child. I argue that this natural approach fails. If you choose to become a parent, and your choice is based on projections about what you think it would be like for you to have a child, your choice is not rational. If you choose to remain childless, and your choice is based upon projections about what you think it would be like for you to have a child, your choice is not rational. This suggests we should reject our ordinary conception of how to make this life-changing decision, and raises general questions about how to rationally approach important life choices.

Paul, L. (2015). What You Can't Expect When You're Expecting Res Philosophica, 92 (2), 1-23 DOI: 10.11612/resphil.2015.92.2.1