Sunday, January 24, 2016

Do we value past experiences?

Derek Parfit's 1984 work, Reasons and Persons, is one of the most important works on moral philosophy. It put forward an account of ethical decision making rooted in a philosophical conception of rationality, identity, and time. It has had a substantial impact on discussions of morals and ethics across a wide range of domains. It is also highly relevant to behavioural economics, in particular in understanding inter-temporal choice. 

The book is divided into 4 sections: the first "Self-defeating theories" deals with rational choice in the context of interconnected decisions; the second "Rationality and Time" develops an account of inter-temporal choice that questions our received notions of time; the third "Personal Identity" develops an account of identity based not on continuity but rather on mental connectedness; the fourth section discusses the nature of obligation to future generations. The book consists of 20 chapters as well as appendices, totalling 154 subsections each containing important ideas. The four parts of the book also interconnect and build to a startling set of ideas on how we should ethically behave towards others, our future selves, and future generations. The book's concepts have substantial relevance for economists. One example is his well-known "repugnant conclusion" and "non-identity problem", one potential consequence being that future generations should not be factored into policy considerations as different courses of action will simply create different future generations who, providing their existence is preferable to them than non-existence, will be grateful to have been born. Most relevant to the type of work we discuss on this blog, Shane Frederick has developed Parfit's ideas on psychological connectedness to future selves in order to examine inter-temporal choice.

Rather than attempt the task of summarising the relevance of Reasons and Persons for behavioural economics, I provide here one example for the purpose of illustration. Chapter 8 "Different Attitudes to Time" examines whether we are neutral with regard to time. If we are valuing a reward or punishment should it matter to us whether we receive it now, in the future, or in the past? A large part of behavioural economics is based on the idea that people will value present rewards more highly than future rewards, and that this will be accentuated when the time horizon is nearer. However, the idea of valuing past rewards is not something that is often discussed. 

At first glance, it seems like a very odd idea. For example, take the choice below (which is my clumsy construction): 

A: You will receive 100 pounds in a week, which you must spend on something you really enjoy. 

B: You will learn that you received 100 pounds last week, which you spent on something you really enjoyed and subsequently forgot about. 

Assuming that both experiences are equally enjoyable, it seems unlikely that anyone would prefer to have had that experience in the past rather than to have it available as a future option. Arguably, this is because we experience time as moving forward from our present state. An experience that happens to us in the past can only be valuable if it provides us with some element of present or future enjoyment. For example, if B led to us meeting someone that we became friends with, it may be rational to prefer B over the prospect of A. But the key point is that we do not place a value on the experience of B in itself because it is a past event that we no longer remember. 

Can we imagine conditions where we value past experiences without regard for any present or future benefits deriving from them? Parfit gives the example of someone learning that their mother had died. Should it matter to the person how she died? Given any information about this refers to something that is in the past, should the person then be indifferent? Consider the case where the person learns that their mother had endured many months of intense pain prior to death and compare this to the case where the person learns that their mother had died painlessly, surrounded by loved ones, after a brief illness. There are many objections one could raise about the precise set-up and I would certainly encourage people to read the very detailed passage by Parfit to fully appreciate the thought experiment. But it seems clear that the valuation of past experience is not just due to future utility considerations, such as the effect on family members and so on. We feel bad that our mother would have suffered and we want to avoid this, even if the event is in the past. Puzzlingly, this probably does not apply to our own past experiences. For example, Parfit discusses the case where you are conscious during a medical operation but then given a drug to make you sleep and forget the experience. On waking up, would it matter to you whether the operation had been painful or not? Compare this to your preferences immediately before the operation. 

The above examples are contained in one chapter that makes a far more wide-ranging point about the nature of time attitudes. This chapter itself is nested in a section that outlines a theory of temporal decision-making and this, in turn, is nested into a full theory of ethical decision making. But, even in isolation, the idea already makes us think about a number of assumptions we rarely question including how we experience time and whether our valuations are necessarily just related to future and present experiences, as opposed to past experiences. Questioning the assumption that we only value present and future experiences can then lead to us to question how we incorporate time into valuation more generally. Parfit goes on to do this in an ingenious way in the rest of the book. 


Anonymous said...


Thanks for posting this. There is enough in this one post to start several different discussions so the fact that all you have done is refer to one of the 145 sub-sections in Parfit's book is striking. I would like to make a couple of brief points.

One is about your comparison between A and B. It seems to me that there are two critical issues here. One is that the person has forgotten what gave her enjoyment. What exactly is the nature of this forgetting? Are we talking about a situation where a person says I know I had a great time in that restaurant but I can't remember what I ate or are we talking about a situation where a person says I was in Italy on holidays in 1986 and I can't remember a thing about what we did nor can I can remember whether I enjoyed it or not.

A related point which I think is often ignored is that some (perhaps many) people continually (perhaps occasionally) reevaluate past experiences (including economic decisions). This is one place where I have long felt that the sunk cost fallacy is not really odd at all. It seems to me very reasonable that we might feel better by being able to justify previous decisions we made in the past and that we find it difficult to let bygones be bygones on those purchasing decisions that did not turn out so well. I think but I am not sure that the philosopher Elizabeth Anderson had some interesting things to say about this in a book she wrote about 20 years ago.

The last point I have is about the issue of whether the way a person dies should matter. I think suicide and other tragic deaths are good examples of how we often recall how a person dies more easily (and with greater emotion) depending on how that death was caused. I have to think about what exactly caused my father's death but would not have such a difficulty with recalling the cause of death of anybody I know who died by suicide or in an accident.

And my very last point is only slightly related but it is something that I have been struck by lately. There is a great research project underway here to elicit QALY weights using the EQ-5D-5L. Part of it involves a time trade-off exercise and a couple of researchers did a (live) demonstration of the method for one of my classed last October. As I listened to the subject being asked would he prefer 10 years of being in imperfect health to X years of being in perfect health it struck me that implicitly in that question is an idea that in the first case death occurs after a long slow painful process whereas in the second death occurs suddenly. I think this may lead to systematic biases such that the utility weights on states of ill health are lower than if the type of death wasn't implied in these trade-offs.

Brendan Kennelly

Liam Delaney said...

Thanks for this Brendan. Several people have pointed out to me that option B is conflating many different things. Obviously the Parfit example of being conscious during an operation but being given a drug to remove all memories is a better example if we want to isolate purely the valuation of the past experience rather than our memory of it. You are correct the value of B might depend on whether we had simply forgotten it but can now recall it, or whether we have no memory of it at all.

I am thinking of how to value negative past life experiences. There are obviously standard utilitarian considerations. Past abuse and unemployment might negatively affect our capacities in various ways that influence our present and future welfare. They may also condition our stress responses to make us evaluate different stimuli in different ways. I think this is already getting close to saying we value the past albeit through its effect on our present emotions. It is interesting to think of cases where we value the past independent of these considerations. The mother example is interesting as it refers purely to an event that is past, that cannot have impinged on us neurologically and is not in our memory. I accept your point that it may condition how we subsequently think of our mother and, in that sense, we are valuing the present and future welfare consequences. But I can't help but think there is an element of tautology in saying that the valuation of the past is due to this concern. It would be worth considering thought experiments where you are told something that you will subsequently immediately have wiped from memory (though the present value would still remain).

From a practical point of view, I think it will be fine to value memories to the extent that they contribute to future flows of well-being and this might be the way to progress for things like scarring effects.

Interestingly, have a look at the list of Bates Clark winners. Most economists will recognise almost all of them, with many going on to win the Nobel prize. They will certainly recognise the first one (Samuelson). They may not recognise the second one, Kenneth Boulding, other than perhaps for his famous phrase that mathematics brought rigour to economics but unfortunately also mortis. While Samuelson developed discounted utility theory, Boulding worked on the idea that life was about accumulating a stock of "psychic capital" stored in memories. The idea did not really catch in Economics but it might be due a revival given the recent developments in behavioural economics.

Liam Delaney said...

Someone sent me the following list of examples of how we might value past events that are unlikely to materially effect us in the present or future. It reads a bit like the contents of a Daily Mail issue, but I think many of these are very useful.

Learning that your husband was a spy.

Learning that you were adopted as a baby but nobody had told you. Now your birth parents are dead but you still have the people you always considered to be your parents.

You consider that you had a very happy childhood. But one day you learn that you were accidentally swapped with another baby in the hospital.

Learning that your partner had cheated on you for years.

Realising that a favourite teacher taught you something fundamentally wrong.

Learning that someone you care about read your diary or personal letters.

Finding out in ten years' time that flossing your teeth has no real health benefit (although it's not harmful).

Finding out years after the fact that you were not the first choice when you were hired.

Finding out that you got your current job despite a previous employer writing a very bad reference for you.

People in Germany finding out years later that their relatives were Nazis or denounced their neighbours during the second world war.

You are now a successful lawyer with a happy family and lots of friends. You meet up with an old friend from school, who tells you that when you were in secondary school, the person you considered your best friend spread horrible rumours about you.

A juror learning that a defendant they found guilty of murder years ago was actually innocent based on DNA evidence (not available at the time of the trial).

Finding out that Santa Claus isn’t real!

Ron said...

Excellent post.
One issue about the experiment cited is that of anticipation.
(A: You will receive 100 pounds in a week, which you must spend on something you really enjoy.
B: You will learn that you received 100 pounds last week, which you spent on something you really enjoyed and subsequently forgot about.)

The future contains expected enjoyment of the experience PLUS anticipation. Anticipation of a positive event is valued. It is possibly linked to the issue of uncertainty – uncertainly is often felt negatively [hence reliable real-time bus times shown at bus stops make waiting for a bus much more pleasant than when you do not know if the bus is coming or not – even though the actual bus arrival time is the same in both cases]. Anticipation can also be positive as in case A above. In other words the A-B choice omits a crucial dimension that I do not think is easy to ‘assume away’ in a meaningful sense.

Liam Delaney said...

Thanks Ron. Yes, I need to disentangle the components of valuation of the future and valuation of the past. It is certainly the case that one reason we value the future is that we value the period between the future and present where we can anticipate what events might happen. Similarly, our valuation of past events is affected by memories of them. I am reading some work on how people remember as opposed to anticipate experiences and there are interesting asymmetries.

With the specific A/B example, I think it is also reasonable to think that people would value A over B even if they were told they would receive A as a complete surprise. It is obviously difficult to think of a thought experiment that would completely remove anticipation, so let me just state for now that I think A is more valuable for B not just because of anticipation but also because people are inclined to see time as moving forward and thus to rate something in the future as inherently more valuable, in these circumstances, than something in the past. The "mother" examples shows that this is not always necessarily the case.

Liam Delaney said...

Thanks to Dave Comerford for the following:

One paper that seemed relevant is Conceptual Consumption by Dan Ariely and Mike Norton. One example is that people engage in uncomfortable experiences today (e.g. a rollercoaster ride; eating pigs' eyes) in order to be able to have the memory and story of having done it.

Another piece that comes to mind is this one from "This American Life". It suggests there are circumstances in which exposing a con can be more costly than leaving it go: