Wednesday, July 25, 2007

"Want" vs. "Should" Decisions - When People Choose the Broccoli over the Baby Ruth

This is a replica of another blog post at, which is more business-oriented, but i think the concept has rather large HCI/d implications, so i'm posting it here as well.

Most people, when making decisions, have to choose between "want" and "should" decisions. But what makes a person choose one over the other?

A recent Harvard Business School Article Understanding the 'Want' vs. 'Should' Decision (link to Harvard Business School's website) helps to answer this. Based on a number of studies of purchasing and voting behavior, as well as some older work in Construal Level Theory, explained nicely in the paper Temporal Construal (link to pdf paper by Trope and Liberman). The studies show that the further away the implementation of a decision is (future or past), the more people will favor a ‘should’ decision over a ‘want’ one.

As a simple example, if someone asks me to decide on and eat immediately either Broccoli or a Baby Ruth candy bar, i am more likely to choose the candy bar, since it's what i want right now. If, on the other hand, a person asks me to decide right now what i will eat at a later time (or if they ask me later what i should have eaten previously), i am more likely to choose the broccoli. What this suggests is that the further away from the present the consequences of my decision are, the more likely i am to choose what i should do.

In the article, the authors state this this "should/want" factor is the reason that most grocery stores place their produce section directly at the front entrance, while the candy bars and trashy magazines appear at the cash registers - to ensure that when walking through the "should" section (healthy produce) shoppers are furthest from the point of consumption (which will probably happen at home, long after they leave the store), and simultaneously to ensure that when entering the "want" section (candy bars), shoppers are the closest to the point of consumption (which most likely will occur in the store or soon after leaving it).

This is an important consideration for HCI/d folks

1) If you are designing a "should" application (something that is more good for the user but less instantly-gratifying), or one that provides the user with "should" features, the further in advance a person makes the decision, the more likely they may be to choose your "should" service over a "want".

2) If you are providing a "want" service, you will need to ensure that your users make the decision as close as possible to the time of consumption - by offering quick setup of their account, express checkout, immediate delivery, etc.

3) It seems to me that offering both "wants" and "shoulds" as a complimentary mix might be a good idea for any design.

4) Since, according to the Trope/Liberman paper, Construal is also related not just to temporal distance, but also spatial, i wonder what this means for distributed communities (i.e., if i am chatting online with a person and can see that they are on the other side of the world, is my experience different than if i know that they are in the same town)

I'll be exploring this further, both in my academic research, and as part of, where we present our users simultaneous "want" and "should" options - that we may want to present differently to maximize their use.

Here is a quote from Trope and Liberman's paper:

"CLT (Construal Level Theory) suggests that temporal distance affects preferences and judgments by changing the way individuals mentally represent future events. The greater the temporal distance from future events, the more likely are the events to be represented in terms of a few abstract and core features (highlevel construals) rather than in terms of more concrete and superficial features (low-level construals). Therefore, temporal distance changes judgments and decisions because in the distant future, compared with the near future, judgments and decisions are more likely to reflect the evaluative and informational implications of high-level construals than those of low-level construals."

No comments: