Perspectives on Psychological Science

5(4) 343–355
General Evaluability Theory ª The Author(s) 2010
Reprints and permission:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/1745691610374586
http://pps.sagepub.com

Christopher K. Hsee1 and Jiao Zhang2

1
Booth School of Business, University of Chicago, Chicago, IL, and 2School of Business Administration, University of Miami,
Miami, FL

Abstract
A central question in psychology and economics is the determination of whether individuals react differently to different values
of a cared-about attribute (e.g., different income levels, different gas prices, and different ambient temperatures). Building on
and significantly extending our earlier work on preference reversals between joint and separate evaluations, we propose a gen-
eral evaluability theory (GET) that specifies when people are value sensitive and when people mispredict their own or others’
value sensitivity. The GET can explain and unify many seemingly unrelated findings, ranging from duration neglect to affective
forecasting errors and can generate many new research directions on topics ranging from temporal discounting to subjective
well-being.

Keywords
evaluation mode, scope sensitivity, duration neglect, preference reversal, temporal discounting

Will people be more willing to take a vacation if the airfare is larger diamond relative to the smaller diamond, the more he
$200 rather than $300? Will people have a more painful mem- is sensitive to diamond size.
ory of a surgery they underwent if the surgery lasted 2 hr rather In what follows, we first introduce the theory, then show how
than 1 hr? Will people work harder to earn $100 now or $120 a it unifies many disparate findings in the existing literature, and
month from now? Will increasing the wealth of all improve the finally suggest new research directions derived from the theory.
happiness of all?
These questions tap a diverse array of topics, ranging from
price and demand, through duration neglect and temporal dis- The Theory
counting, to wealth and happiness. Yet they all concern the
Overview
same fundamental issue: value sensitivity.
The topic of value sensitivity is essential to both psychology Our theory, which we call the general evaluability theory
and economics. Psychophysics (concerning how one reacts (GET), seeks to address three broad questions: What factors
to sensory input), prospect theory (concerning how one determine value sensitivity? What is the relationship between
values outcome and weighs probability), theory about price these factors? Can people accurately predict value sensitivity?
elasticity of demand (concerning how demand changes as price The GET starts from three basic propositions, each addressing
changes), and happiness research (concerning how happiness is one of the questions, and generates an array of specific hypoth-
related to income and other variables) are all about value eses. The chart in Figure 1 offers an overview.
sensitivity. We should note here that GET greatly extends previous work
This article presents an integrative theory about value sensi- on evaluability (e.g., Hsee 1996, 1998; Hsee, Blount, Loewen-
tivity—namely, about when one’s evaluation is responsive to stein, & Bazerman, 1999). First, the notion of evaluability is now
variations in value and when it is not. Value is a level of an greatly expanded to three different types. Second, although the
objective attribute one cares about, such as price, duration, previous research on evaluability focused mainly on preference
probability, wealth, and so on, and evaluation is one’s subjec- reversals, the present theory encompasses a much wider range of
tive reaction, including willingness to purchase, willingness to
work, happiness, and so on. If a person is willing to pay as
Corresponding Author:
much for a 1-karat diamond as he would for a 0.5-karat dia- Christopher K. Hsee, Booth School of Business, University of Chicago, 5807 S.
mond, we say that the person is insensitive to diamond size Woodlawn Avenue, Chicago, IL 60637
(value) in that range. The more he is willing to pay for the E-mail: [email protected]

343
344 Hsee and Zhang

Proposition 1 Basic Effects Ariely et al., 2003; Hsee &

Zhang, 2004; Kogut & Ritov,
Evaluability depends H1.1: Mode × Value
2005; Zhang & Hsee, 2009
on mode, knowledge, H1.2: Knowledge × Value
and nature
H1.3: Nature × Value
Ariely & Loewenstein, 2000;
Hsee & Shen, 2009; Morewedge
Hsee et al., 2009
et al., 2009; Yeung & Soman, 2007;
Zhang & Hsee, 2009

Moderated Basic Effects Hsee & Shen, 2009;

H2.1: Mode × Value × Knowledge Zhang & Hsee, 2009
H2.2: Mode × Value × Nature
Proposition 2 Hsee et al., 2009
H2.3: Knowledge × Value × Nature
Mode, knowledge, and
nature are conjunctive
to produce low Bazerman et al., 1992;
sensitivity JE–SE Preference Reversals Cryder et al., 2008; Hsee,
H3.1: Mode × Option (knowledge) 1996, 1998; Kogut &
Ritov, 2004; List, 2002;
H3.2: Mode × Option (nature)
Lowenthal, 1993; Paharia
et al., 2009; Schmeltzer
et al., 2004; Willemsen &
Keren, 2004; Zikmund-
Basic Mispredictions Fisher et al, 2004
H4.1: Role (mode) × Value
H4.2: Role (knowledge) × Value
Dunn et al., 2003;
Hsee & Zhang, 2004
Proposition 3 Moderated Mispredictions
People in a high H5.1: Role (mode) × Value × Knowledge
evaluability condition
H5.2: Role (mode) × Value × Nature
will overpredict value
sensitivity of those in a H5.3: Role (knowledge) × Value × Mode
low evaluability H5.4: Role (knowledge) × Value × Nature
condition

JE–SE Mispredictions
H6.1: Role (mode) × Option (knowledge) Hsee & Zhang,
2004
H6.2: Role (mode) × Option (nature)

Fig. 1. Propositions and hypotheses of the general evaluability theory. JE ¼ joint evaluation; SE ¼ single evaluation.

topics, rendering preference reversals only an element of the cur- Proposition 1 (Evaluability)
rent framework. Third, the current theory integrates numerous
We now introduce the three basic propositions of the GET and
new findings that have emerged since the publications of the
their corresponding hypotheses. The central tenet of our theory
original papers on evaluability in the late 1990s. Finally, the cur-
is that value sensitivity depends on evaluability. Evaluability
rent theory generates several new research directions which we
refers to the extent to which a person has relevant reference
believe are general and significant.

344
General Evaluability Theory 345

Table 1. Definitions of Evaluability Levels for Each of the Three being equal, diamond traders (with rich knowledge) will have
Factors a stronger reaction toward a 0.5-karat diamond versus a 1-
Factor Low-evaluability level High-evaluability level karat diamond than would regular consumers (with poor
knowledge).
Mode Single evaluation (SE) Joint evaluation (JE) A general corollary of H1.2 is that, everything else being
Knowledge Poor Rich equal, evaluability is higher for values that constitute categori-
Nature Inherently inevaluable Inherently evaluable cal differences rather than incremental differences. Two values
are said to constitute a categorical difference if they differ in
information to gauge the desirability of target values and map sign or in type, such as whether a stock price rises or falls, or
them onto evaluation. For instance, college students have abun- whether an apartment features indoor parking or not. Two val-
dant information about grade point average (GPA), such as its ues are said to constitute an incremental difference if they dif-
mean and range, but little knowledge about body-mass index fer only in degree, such as how much the stock price rises
(BMI, a measurement of height-adjusted body weight). Thus, today, or how large the apartment is. Generally speaking, peo-
for college students, evaluability of GPA is high and evaluabil- ple are more knowledgeable about sign and type than about
ity of BMI is low.1 degree—for example, most people know that a positive invest-
Our first proposition concerns the determinants of ment return is good and a negative return is bad, but would not
evaluability: know as clearly how good a given positive return is or how bad
a given negative return is. Thus, categorical differences lead to
Proposition 1: Evaluability of, and hence value sensitivity to, greater value sensitivity than do incremental differences. More
values depends on three factors: mode, knowledge, and nature. precisely, people in SE are more sensitive to categorical differ-
ences than to incremental differences, holding their sensitivity
Each of the three factors can assume either a low- or high- to the two types of differences in JE constant. (See our discus-
evaluability level, as summarized in Table 1. For each factor, sion of the Mode  Value  Knowledge interaction later in this
the low-evaluability level produces low-sensitivity evaluations article.)2
and the high-evaluability level produces high-sensitivity eva- Nature (H1.3). Nature refers to whether human beings have
luations, thus yielding a Factor  Value interaction (see the left an innate and stable physiological or psychological ‘‘scale’’
panel of Figure 2—the figure is only illustrative; we do not (reference system) to evaluate values on an attribute. The attri-
assume that the slope for the low-evaluability level is entirely bute is inherently evaluable if they do or inherently inevaluable
flat or that there is no main effect between the high- and the if they do not. Ambient temperature is an example of an inher-
low-evaluability levels.) We now elaborate on each factor in ently evaluable attribute; even without learning or social com-
turn. parison, we can tell what temperature makes us comfortable
and happy and what does not. Other examples include amount
Mode (H1.1). Any evaluation is formed in one of two modes
of sleep, social isolation, or connectedness. The size of a dia-
or some combination of the two: the joint evaluation (JE)
mond and the power of a car are examples of inherently ineva-
mode, in which two or more values are juxtaposed and com-
luable attributes; without learning or comparison, we would not
pared, or the single evaluation (SE) mode, in which only a sin-
know how to assess such variables. Of course, some people
gle value is presented and evaluated. Because evaluators can
know how to evaluate diamond size and car power, but such
use one value as a reference to evaluate another value in JE, but
knowledge is learned, not innate. Because people possess
not in SE, evaluability, and hence value sensitivity, will be
innate reference systems for inherently evaluable attributes but
higher in JE than in SE (H1.1). For example, individuals will
not for inherently inevaluable attributes, value sensitivity
be more sensitive to the size of a diamond in JE than they
(without learning or comparison) is higher for inherently evalu-
would in SE.
able attributes (H1.3). More precisely, people in SE are more
Knowledge (H1.2). Knowledge refers to the amount of distri- sensitive to differences on an inherently evaluable attribute
butional information (e.g., the range, the mean) about the target than to differences on an inherently inevaluable attribute, hold-
attribute one has acquired through learning. Knowledge may ing their sensitivity to the two types of differences in JE con-
vary across individuals for the same attribute or across attri- stant; see our discussion of the Mode  Value  Nature
butes for the same individual. Diamond traders know more interaction later in this article.
about the sizes of diamonds than do ordinary consumers— an It should be noted that classifying a variable as inherently
example of knowledge varying across individuals for the same evaluable does not mean that it is immune to the influence of
attribute. College students know more about GPA than about external reference information (such as social comparison);
BMI—an example of knowledge varying across attributes for instead, it means that people can evaluate the variable even
the same individuals. without such information. Also, inherently evaluable variables
Evaluators have more reference information about an attri- are not always associated with basic biological needs—they
bute if they have rich knowledge about the attribute. Thus, also include socio-psychological variables, such as loneliness,
richer knowledge engenders higher evaluability and, hence, depression, and sense of achievement. (For details, see Hsee,
higher value sensitivity (H1.2). For example, everything else Yang, Li, & Shen, 2009.)

345
346 Hsee and Zhang

Low-Evaluability Level of High-Evaluability Level of

Another Factor Another Factor

Evaluation

worse value better value worse value better value

high-evaluability level of factor

low-evaluability level of factor

Fig. 2. Evaluability levels between factors. The left panel shows a basic Factor  Value interaction,
where factor (the primary factor) can be mode, knowledge, or nature. The right panel shows that
the basic two-way interaction can be moderated by another factor (a moderator), where the
moderator can be any of the three factors other than the one in the basic two-way interaction.

All about reference information. Mode, knowledge, and nature That is, the final evaluation will be of low sensitivity only if all
are three sources of reference information. Mode is the most three factors are of low evaluability.
‘‘here and now’’ source, concerning reference information that
resides in the set of values or options presented; knowledge is Although the conditions producing low sensitivity appear strin-
less here and now, concerning reference information acquired gent (requiring SE, poor knowledge, and inherent inevaluabil-
through past experience; and nature is the least here and now ity), they are actually quite common: We often experience
source, concerning reference information that is innate, hard- events one at a time, our knowledge is limited, and many vari-
wired, and rooted in evolution. Put differently, the three factors ables we pursue accord no innate evaluation scales. For exam-
represent three types of evaluability: Mode is ad hoc evaluabil- ple, most of us live in only one home, know little about how to
ity, knowledge is learned evaluability, and nature is inherent translate size into desirability, and have no innate scales to
evaluability. The distinction between learned evaluability evaluate the size.
(knowledge) and inherent evaluability (nature) is not an artifi- Proposition 2 generates two sets of hypotheses: one about
cial classification, and it has important implications, which we moderators for the three basic two-way interactions and one
will explore later. about JE–SE preference reversals involving multiattribute
Mode, knowledge, and nature are three independent dimen- options. We discuss these hypotheses in turn.
sions and none of them can subsume another. Just as any phys-
ical object has three dimensions, length, height, and width, any Moderated effects (H2.1–H2.3). Because the three factors
evaluation assumes a value on each of the three factors. The are conjunctive in producing low value sensitivity, any one
evaluation is of either inherently evaluable or inherently ineva- of the three basic Factor  Value interactions (H1.1–H1.3) can
luable attributes, by evaluators with rich or poor knowledge, be moderated by another factor, as depicted in Figure 2. This
and in JE or SE. yields the 3 three-way interactions described in H2.1–H2.3.
The Model  Value  Knowledge interaction predicts, for
example, that individuals are more sensitive to diamond size in
JE than in SE, but this effect is more pronounced if these indi-
Proposition 2: Relationship Among the Three viduals are diamond novices rather than diamond experts. It
Factors also predicts that people are more sensitive to the size of an
According to our theory, value sensitivity will be high as long apartment (an incremental difference) in JE than in SE, but
as the evaluator has the relevant reference information, regard- their sensitivity to whether the apartment has indoor parking
less of whether the reference information comes from mode, (a categorical difference) is relatively similar in JE and SE. Put
knowledge, or nature. Thus, we propose the following: differently, holding sensitivity to size and parking constant in
JE, people in SE are more sensitive to parking than to size.
Proposition 2 (relationship among the factors): Mode, knowl- The Mode  Value  Nature interaction predicts that peo-
edge, and nature are conjunctive in producing low sensitivity. ple are more sensitive to diamond size in JE than in SE, but

346
General Evaluability Theory 347

they will be relatively equally sensitive to temperature in JE predictors know that the predictees are in a low-evaluability
and in SE. Finally, the Knowledge  Value  Nature interac- condition.
tion proposes that learned knowledge about jewelry can greatly
increase one’s sensitivity to diamond size, but learned knowl- Note that the reverse is not applicable, namely, predictors in a
edge about temperature will not exert such a big effect on one’s low-evaluability condition (e.g., SE) will not mispredict the
sensitivity to temperature, because organisms are already value sensitivity of predictees in a high-evaluability condition
sensitive to temperature without external knowledge. (e.g., JE). The reason is that to make a prediction for predictees
JE–SE preference reversals (H3.1–H3.2). All the hypotheses in a high-evaluability condition, predictors will have to be
introduced so far concern evaluations of values on a single informed of the additional reference information (e.g., the
attribute. What about evaluations of two options that involve alternative option in JE) in the high-evaluability condition, and
a tradeoff along two attributes? According to Proposition 2, once they have this additional information, they are effectively
preferences for the two options may reverse between JE and in the same condition as the predictees.
SE if one attribute is already of high evaluability in SE and Recall that Propositions 1 and 2 yield three sets of hypoth-
there is little room for JE to bolster its evaluability, whereas the eses: basic effects, moderated effects, and JE–SE preference
other attribute is of low evaluability in SE but JE considerably reversals. Likewise, Proposition 3 yields three sets of hypoth-
bolsters its evaluability. eses: basic mispredictions, moderated mispredictions, and
There are two types of such JE–SE preference reversals: one JE–SE mispredictions.
involving knowledge (H3.1) and one involving nature (H3.2). Basic mispredictions (H4.1–H4.2). Theoretically, mispredic-
The first constitutes a Mode (JE vs. SE)  Option (poor knowl- tion may happen if predictors are in the high-evaluability level
edge vs. rich knowledge attribute tradeoff) interaction, and the and predictees are in the low-evaluability level on any of the
second is a Mode (JE vs. SE)  Option (inherently evaluable three factors. However, no mispredictions can be made about
vs. inherently inevaluable attribute tradeoff) interaction. nature because nature concerns different types of attributes.
Consider an early demonstration of the first type: The study Thus, Proposition 3 yields only two basic mispredictions: one
(Hsee, 1996) involved evaluations of two job candidates for a about mode and one about knowledge. The first posits that pre-
computer programmer position, one with a higher GPA and dictors in JE will overpredict the value sensitivity of predictees
the other with more programming experience. In SE, the in SE, which implies a Role (mode)  Value interaction (H4.1).
candidate with the higher GPA was favored, whereas in JE, The second posits that predictors with rich knowledge will
the candidate with more experience was favored. Presumably, overpredict the value sensitivity of predictees with poor knowl-
the participants (all college students) possessed richer edge, which implies a Role (knowledge)  Value interaction
knowledge about GPA than about programming experience. (H4.2). Note that in each case, predictors always assume the
high-evaluability level of the factor and predictees always
assume the low-evaluability level.
Proposition 3: Mispredictions Moderated mispredictions (H5.1–H5.4). In the section detail-
This proposition concerns prediction of one’s own or another’s ing Proposition 2, we explained that any of the basic
value sensitivity. Technically, it is independent of the first two Factor  Value interactions could be moderated by another
propositions, but incorporating it enriches our theory and factor. Likewise, either basic misprediction can be moderated
allows us to explain more existing findings and generate more by another factor, yielding four possible combinations, as
meaningful hypotheses. described in H5.1–H5.4.3 Each of these hypotheses is ecologi-
Ample evidence indicates that if predictors and predictees cally meaningful and testable—take H5.1 for example. Sup-
(who can be either the predictors themselves at a future time pose that a man is in a jewelry store shopping for a diamond
or other individuals) are in different visceral states, predictors ring for his fiancée and is comparing diamonds of different
will project their own feelings onto the predictees and mispre- sizes. Chances are that he will overpredict his fiancée’s sensi-
dict the predictees’ feelings (e.g., Gilbert, Gill, & Wilson, tivity to the size of the diamond and spend too much on a big
2002; Loewenstein, O’Donoghue, & Rabin, 2003; Van Boven, diamond because he is in JE when buying the diamond whereas
Loewenstein, & Dunning, 2005). For example, hungry diners she is in SE upon receiving it (H4.1). However, this mispredic-
tend to overestimate how much they will enjoy a rich dessert tion will be less pronounced if they (especially she) are knowl-
after a rich dinner even though they (should) know that they edgeable about diamonds (e.g., she was married and had
will be full after the dinner. received diamond rings before) than it would be if she is not
By the same token, if predictors and predictees are in knowledgeable.
different evaluability conditions, misprediction will also arise. JE–SE mispredictions (H6.1–H6.2). Just as JE–SE preference
Specifically, reversals may arise if two options vary along a high-
evaluability attribute and a low-evaluability attribute, so will
Proposition 3: If predictors are in a high-evaluability condition JE–SE mispredictions occur for such options if the predictor
and predictees (either others or predictors themselves in the is in JE and the predictee in SE. Specifically, there are two pos-
future) are in a low-evaluability condition, the predictors will sible JE–SE mispredictions, one involving knowledge (H6.1)
overestimate the predictees’ value sensitivity, even if the and one involving nature (H6.2). To illustrate, H6.2 posits that

347
348 Hsee and Zhang

if two options involve a trade-off between an inherently- (2000) found that participants were sensitive to the durations of
evaluable attribute and an inherently-inevaluable attribute, and noises if they compared the noises with a referent noise.
if predictors are in JE while predictees are in SE, then the More generally, we propose that the intensity of an event is
predictors may mistakenly predict the predictees to favor the inherently easier to evaluate than the duration of the event.
option superior on the inherently inevaluable attribute. Humans have an innate psychophysical scale to tell how loud
a noise feels aversive and how strong an electrical shock is
painful, but do not have such an inherent reference system
Integrating Existing Findings to evaluate their durations. Most individuals would find a
The GET can parsimoniously explain various seemingly unre- 100-dB noise aversive but would not have much idea whether
lated findings. Below are some examples. a 100-s noise was long or short. That is probably why we often
observe sensitivity to intensity and insensitivity to duration.
Scope Insensitivity and Duration Neglect
JE–SE Preference Reversals and Arbitrary
People are sometimes clueless about the ‘‘scope’’ (e.g., fre-
quency, size, or duration) of an event or outcome that they care
Cohesiveness
about. People are willing to pay as much to save 2,000 birds as Intuitively, two items (e.g., two job candidates) evaluated com-
they would to save 200,000 birds (Desvousges, Johnson, paratively (i.e., in JE) or separately (i.e., in SE) should yield the
Dunford, Boyle, & Wilson, 1993). They are willing to donate same preference ranking. If Candidate A is rated more favor-
as much to help a single victim as they would to help a group ably than Candidate B in JE, then A should also be rated more
of victims (Kogut & Ritov, 2005). However, these dramatic favorably than B in SE. In reality, preference ranking often
scope-insensitivity effects occur only when different values reverses between JE and SE (e.g., Bazerman, Loewenstein, &
(e.g., 2,000 birds vs. 200,000 birds) are presented in SE, and Blount White, 1992; Bazerman, Schroth, Shah, Diekmann, &
they disappear when they are presented in JE. In our opinion, Tenbrunsel, 1994; Bazerman, Tenbrunsel, & Wade-Benozi,
these results are manifestations of the Mode  Value interac- 1998; Chatterjee, Heath, & Min, 2009; Cryder, Mullen, &
tion in H1.1. Loewenstein, 2008; Gonzalez-Vallejo & Moran, 2001; Hsee,
The most studied form of scope insensitivity is duration 1996, 1998; Hsee et al., 1999; Irwin, Slovic, Lichenstein, &
neglect, which states that retrospective evaluation of an event McClelland, 1993; Kahneman & Ritov, 1994; Kogut & Ritov,
is insensitive to its duration (e.g., Fredrickson & Kahneman, 2005; List, 2002; Loewenthal, 1993; Moore, 1999; Paharia,
1993; Kahneman, Fredrickson, Schreiber, & Redelmeier, Kassam, Greene, & Bazerman, 2009; Schkade & Kahneman,
1993). In a well-known experiment (Kahneman et al., 1993), 1998; Schmeltzer, Caverni, & Warglien, 2004; Tehila & Ritov,
respondents experienced two episodes one at a time (in SE 2005; Willemsen & Keren, 2004; Zikmund-Fisher, Fagerlin, &
though within-participants). (By default, temporally extended Ubel, 2004; see Larrick, 2004, for a discussion on how to mini-
events are experienced one at a time, as in SE or quasi-SE.) One mize these and other biases).
episode (the shorter one) required participants to immerse their For example, Zikmund-Fisher et al. (2004) reported that
hands in very cold water for 60 s; the other episode (the longer patients in SE favored less experienced surgeons who gradu-
one) required them to immerse their hands in very cold water ated from Harvard over more experienced surgeons who grad-
for 60 s and then in mildly cold water for another 30 s. Objec- uated from the University of Iowa, though in JE their
tively the longer episode was worse, but respondents remem- preference were reversed. Presumably, patients had more
bered it as being less aversive and were more willing to knowledge about the quality of universities than about the
repeat it, indicating sensitivity to the final intensity (very cold experience of a doctor. This, like many other such findings,
or mildly cold) of the experience but insensitivity to the dura- is a manifestation of the Mode  Option (knowledge) interac-
tion (60 s or 90 s). tion in H3.1.
We suggest that a possible reason for duration neglect is that As mentioned earlier, a special case of the JE–SE preference
duration is inherently inevaluable and it is also low in learned reversals in H3.1 arises when the options to be evaluated
evaluability—in most experiments showing duration neglect, involve an incremental difference on one attribute and a cate-
the stimuli are unfamiliar to the participants. Recently, More- gorical difference on another. For example, List (2002) exam-
wedge, Kassam, Hsee, and Caruso (2009) asked respondents ined bidding prices for two sets of baseball cards, one
to retrospectively evaluate a noise, which was either unlabeled containing 10 cards, all of which were described as being in
or labeled as an Australian phone ring, that lasted either 7 s or mint condition, and the other containing 13 cards, of which
21 s. They replicated the typical duration neglect finding in the 10 were described as being in mint condition and 3 were
no-label condition but observed significant duration sensitivity described as being in poor condition. In SE, bidding was higher
in the labeled condition. Presumably the labeling prompted for the pack of 10 mint-condition cards; in JE, bidding was
respondents to use the duration of a usual (familiar) phone ring higher for the pack of 13 mixed-condition cards. Notice that
as a reference, turning a poor-knowledge experience into a whether a set included 10 or 13 cards was merely an incremen-
rich-knowledge experience. This result reveals the Knowledge tal difference, and whether the set contained all mint-condition
 Value interaction in H1.2. Similarly, Ariely and Loewenstein cards or contained some poor-condition cards was a categorical

348
General Evaluability Theory 349

difference. The baseball card study was a replication of a series psychology and decision theory (e.g., Gilbert et al., 2002;
of original studies by Hsee (1998) showing the same type of Gilbert, Pinel, Wilson, Blumberg, & Wheatley, 1998; Kahne-
preference reversals. In one study, participants in SE were will- man & Snell, 1992; Loewenstein et al., 2003; Wilson &
ing to pay more for a 24-piece intact dinnerware set than for a Gilbert, 2003).
40-piece dinnerware set that contained a few broken pieces, yet A general type of forecasting error is the distinction bias:
their preferences reversed in JE. In another study, participants People faced with multiple options (i.e., in JE) overpredict the
in SE were willing to pay more for 7 oz of ice cream served in value sensitivity of their future selves or of others who experi-
a 5-oz cup (overfilled) than for 8 oz of ice cream served in a ence only one of the options (i.e., in SE; Hsee & Zhang, 2004).
10-oz cup (underfilled), but their preference reversed in JE. For example, Dunn, Wilson, and Gilbert (2003) observed that
More such JE–SE reversals have been reported recently. before finding out which of 12 dormitory buildings they would
For example, Paharia et al. (2009) reported that participants be assigned to, college freshmen predicted that the physical
in SE judged companies that caused direct and less severe features of the buildings would have a great impact on their
harm to consumers more negatively than companies that experience when they lived in each building; however, after
caused indirect (through an agent) but more severe harm, they had been living in the buildings assigned to them, the
whereas in JE their judgment was reversed. Cryder et al. physical features had little impact. One possible explanation
(2008) found that a hedonic but less expensive reward (e.g., of this intriguing finding is that the freshmen were in JE when
$1.49 chocolate) was more effective in SE than a nonhedonic making the predictions and in SE when living in the assigned
but more expensive reward (e.g., $3 cash) in motivating par- dorm and that they were more sensitive to physical variations
ticipants to complete a survey, whereas in JE the latter reward in JE than in SE, which is a manifestation of the Role (mode)
was more effective. The authors explained their findings in  Value interaction in H5.1. The misprediction in this study
terms of a conflict between what participants wanted and what may also result from other factors, such as failure to predict
they thought they should choose, but these findings are also hedonic adaption and failure to consider social-life differences
consistent with our hypothesis (H3.1). In Paharia et al.’s between the buildings.
study, whether the harm was direct or through a third party In a more controlled test of the distinction bias (Hsee &
was a categorical difference, whereas the difference in the Zhang, 2004), participants were assigned to be either predictors
severity of the harm was incremental. In Cryder et al.’s study, or experiencers. The predictors were shown four lists of words,
whether the reward was hedonic or not was categorical, which varied both in valence (positive or negative words) and in
whereas the difference in the worth of the reward was incre- length (10 or 25 words), and were asked to predict how a person
mental. Indeed, most, if not all, of the JE–SE preference would feel after reading only one of the lists. Each experiencer
reversals documented in the existing literature can be parsi- read only one of the lists and reported feelings. Notice that the
moniously explained by the GET. predictors were in JE and the experiencers were in SE. The
Another general phenomenon that seems to be consistent results matched the Role (mode)  Value  Knowledge interac-
with the GET is what Ariely, Loewenstein, and Prelec tion in H5.2: The predictors thought that SE experiencers would
(2003) call ‘‘coherent arbitrariness.’’ In an ingenious study, be sensitive to both the valence of the lists (a categorical differ-
the authors first had respondents listen to a 30-s noise, then ence) and the length of the lists (an incremental difference), yet
asked if they were willing to hear the same noise again for SE experiencers were sensitive only to valence, not to length.
$.10 or $.50 (between-subjects manipulation), and finally Mispredictions can breed mischoices. Choice is typically
asked how much they would demand to hear longer and made in JE, whereas the consequence of the choice is often
shorter noises (within-subjects manipulation). Participants’ experienced in SE. For example, when buying a sofa, we usu-
demands were heavily influenced by the arbitrary anchor, ally compare multiple options, whereas when using the sofa we
demanding more in the $.50 condition than in the $.10 condi- have eventually bought, we face that sofa alone. According to
tion, yet were coherent with the durations of the noises—the H6.1, when two options involve a tradeoff between a rich-
participants demanded more money for longer noises. These knowledge attribute and a poor-knowledge attribute, predictors
findings are consistent with our hypothesis that people exhibit in JE are more likely to favor the option superior on the poor-
low value sensitivity in SE and high value sensitivity in JE. knowledge attribute than will experiencers in SE. Similarly,
Notice that the anchor ($.10 or $.50) was manipulated according to H6.2, when two options involve a trade-off
between subjects (SE) and the durations were manipulated between an inherently evaluable attribute and an inherently
within subjects (JE). Thus, participants in SE were insensitive inevaluable attribute, predictors in JE are more likely to favor
to how much they should be paid and so were influenced by the option superior on the inherently inevaluable attribute. To
the anchor but they were sensitive to duration variations in the extent that people resolve choices by relying on their pre-
JE and therefore demanded more for longer noises. dictions, mispredictions as described in H6.1 and H6.2 will
result in mischoices.
Such a mischoice is demonstrated in Hsee and Zhang
Misprediction and Mischoice (2004). Participants in the study were asked to choose between
Biases in affective forecasting (prediction of future hedonic two tasks: one asking them to recall a sad personal story and eat
experiences) have been the focus of a large literature in social a large piece of chocolate and the other to recall a happy

349
350 Hsee and Zhang

A B C

Fig. 3. Value (A), probability weighting (B), and temporal discounting (C) functions. For all three functions, the curve is
more linear in joint evaluation (solid line) than in single evaluation (dotted line).

personal story and eat a small piece of chocolate. The result (within cities). These findings reflect the Mode  Value 
revealed a stark choice–experience inconsistency: When mak- Nature interaction in H2.2, where room temperature is inher-
ing the choice, most participants chose the sad-story/large- ently evaluable and jewelry value is inherently inevaluable.
chocolate task, but when doing the chosen task, participants The findings help answer the question of whether happiness
who chose the happy-story/small-chocolate task were happier depends only on relative consumption levels or also on absolute
than were those who chose the sad-story/large-chocolate task. consumption levels: For inherently inevaluable attributes such
This pattern is consistent with H6.1: The difference in the size as the monetary worth of possessed jewelries, only relative
of a piece of chocolate is incremental and the difference in the consumption levels influence happiness—that is, happiness
nature of the story (sad vs. happy) is categorical. depends only on social comparison. In contrast, for inherently
evaluable attributes, such as temperature, both relative and
absolute consumption levels influence happiness.
Relative and Absolute Determinants of
Happiness Directions for Future Research
Over the past decades, scholars have vigorously argued about To illustrate new research directions that the GET can generate,
whether happiness is relative or absolute. Some (e.g., Easterlin, we discuss three overarching topics: valuation, temporal dis-
1974, 1995) believe that happiness depends only on relative counting and probability weighting; price and demand; and
outcomes, such as social comparison (Festinger, 1954); for wealth and happiness.
example, how happy people feel about their incomes depends
only on how their incomes compare with their neighbors’, their
Valuation, Temporal Discounting, and
friends’, and their colleagues’. Others (e.g., Veenhoven, 1991)
believe that happiness depends on absolute outcomes and does
Probability Weighting
not rely on social comparison. Which view is correct? A recent Arguably the three most important classes of functions in judg-
study (Hsee et al., 2009) suggests that the answer depends on ment and choice are value (utility) functions, temporal dis-
the nature of the variable. counting functions, and probability weighting functions.
In the study, residents in 31 representative cities in China Value functions describe how objective quantities map onto
were approached during the winter in a large telephone survey subjective evaluations (e.g., Kahneman & Tversky, 1979), tem-
and asked about their room temperature, the monetary worth of poral discounting functions describe how decision makers dis-
their jewelry, and their happiness with these variables. The count the values of temporally delayed outcomes (e.g.,
authors analyzed the effects of temperature and jewelry value Fredrick, Loewenstein, & O’Donoghue, 2002; Read, 2004;
on happiness both within cities and across cities, observing that Soman et al., 2005), and the probability weighting functions
residents in China would compare with each other within a city describe how probabilities translate into decision weights
but not across cities, and hence, within-city evaluations (e.g., Kahneman & Tversky, 1979).
mimicked JE and between-city evaluations mimicked SE. The These three types of functions share a common characteris-
result for temperature emerged to be highly different from that tic: They are all nonlinear, especially around endpoints—zero
for jewelry value. Temperature had both within-city and quantity in value, no delay in temporal discounting, and impos-
between-city effects, but jewelry value had only a within-city sibility (0%) or certainty (100%) in probability weighting
effect, suggesting that temperature could influence happiness (Fig. 3). A large body of research has examined factors that
even without social comparison (between cities) and jewelry influence the nonlinearity of each of the three types of
influenced happiness only when social comparison was easy functions (e.g., Gonzalez & Wu, 1999; Hsee & Rottenstreich,

350
General Evaluability Theory 351

2004; Kluger, Stephan, Ganzach, & Hershkovitz, 2004; Read, between value, delay, and/or probability. For example, con-
Fredrick, Orsel, & Rahman, 2005; Rottenstreich & Hsee, sider two options: A ¼ receiving $100 guaranteed and B ¼ a
2001). 90% chance at receiving $150. On the basis of H3.1, we expect
We now propose another factor that influences the nonli- A to be favored in SE and B to be favored in JE, because the
nearity of all of the three types of functions: evaluation mode. difference in probability (100% vs. 90%) is categorical and the
All of these functions are less linear in SE than in JE. difference in payoff ($100 vs. $150) is incremental. Similarly,
Value functions. We propose that value functions are more consider another pair of options: C ¼ receiving $200 now, and
linear in JE than in SE, as depicted in the left panel of Figure 3. D ¼ receiving $250 in 3 weeks. We expect C to be favored in
In JE, people will be happier with $10 than with $0 or with $20 SE and D to be favored in JE, because the difference in delay
than with $10, and the two differences in happiness will be (no delay vs. a 3-week delay) is categorical and the difference
similar. In SE, people will still be happier with $10 than with in payoff ($200 vs. $250) is incremental.
$0, but will be similarly happy with $20 and with $10. The dif- To the best of our knowledge, we are the first to suggest that
ference between nothing ($0) and something (e.g., $10) is cate- mode influences the shapes of all three types of functions.
gorical, whereas the difference between something small and Future research testing the predictions delineated in this section
something larger (e.g., $10 vs. $20) is incremental. According may yield fruitful results.
to H2.1, people in JE will be similarly sensitive to the incre-
mental difference and to the categorical difference, and people
in SE will be sensitive to the categorical difference but insen-
Price and Demand
sitive to the incremental difference. We have evidence for this The relationship between price and demand is a fundamental
prediction (Hsee & Zhang, 2004; Zhang & Hsee, 2009). In a issue in both microeconomics and marketing. It is generally
recent study (Zhang & Hsee, 2009), participants in JE evalu- believed that the higher the price of a good, the lower the
ated payoffs of different magnitudes (0 chocolates, 5 choco- demand for it, holding supply constant. The sensitivity of
lates, and 10 chocolates) rather linearly, whereas those in SE demand to price is called price elasticity. Economists and mar-
evaluated 0 chocolates and 5 chocolates quite differently but keters have identified various determinants of price elasticity,
evaluated 5 chocolates and 10 chocolates similarly. such as necessity of the good and the wealth of the consumer.
Probability weighting functions. Just as value functions are less Besides these factors, our theory suggests another determinant
linear in SE than in JE, we propose that probability weighting of price elasticity: price evaluability.
functions are also less linear in SE than in JE, as depicted in the The price evaluability of a good depends on whether consu-
middle panel of Figure 3. The reason is similar: Impossibility mers are knowledgeable about the price of that good. For
(0%) versus some possibility is a categorical difference, as is example, most city dwellers are more knowledgeable about
some possibility versus certainty (100%), but a small possibil- taxi fares than about horse-ride fares. Thus, according to our
ity versus a larger possibility is only an incremental difference. theory (H1.2), price elasticity will be higher for taxi rides than
According to H2.1, we predict that the probability weighting for horse rides. The finding of a recent study (Study 1 in Hsee
function will be more S-shaped in SE than in JE. For example, & Shen, 2009) is consistent with this prediction. In the study,
in JE, one will be happier with $100 guaranteed than with a participants were either asked to indicate whether they would
90% chance at $100, and happier with a 90% chance at $100 take a taxi or a horse ride to a destination. The price for the ride,
than with an 80% chance at $100. In SE, however, one will low or high, was manipulated between participants. A pretest
be happier with $100 guaranteed than with a 90% chance at confirmed that participants were knowledgeable about taxi
$100, but will feel similarly happy with a 90% chance at fares but unknowledgeable about horse-ride fares. As pre-
$100 and with an 80% chance at $100. dicted, willingness to take a taxi was highly sensitive to the
Temporal discounting functions. As with the value and the price of the taxi, whereas willingness to take a horse ride was
probability weighting functions, we propose that the temporal insensitive to the price of the horse ride.
discounting functions are also less linear in SE than in JE, as Can marketers and sellers accurately predict buyers’ price
shown in the right panel of Figure 3. The difference between elasticity? There is reason to believe that marketers and sellers
no delay and some delay is categorical, whereas the difference are likely to overpredict buyers’ price elasticity. When setting
between a short delay and a longer delay is merely incremental. the price for a good, sellers typically entertain different prices
According to H2.1, we predict that around the zero (no delay) before settling on a price. Potential buyers of the good, on the
point, the temporal discounting function will be more convex in other hand, only see the price the sellers have set. In other
SE than in JE. For example, in JE one will feel happier with words, the sellers are in JE and the buyers are in SE. According
$100 now than with $100 in 2 weeks, and happier with $100 to our theory (H6.1), sellers will overpredict buyers’ price elas-
in 2 weeks than with $100 in 4 weeks. In SE, however, one will ticity if buyers have poor knowledge about the price, but not if
feel happier with $100 now than with $100 in 2 weeks, but will buyers have rich knowledge about the price. Overpredicting
feel similarly happy with $100 in 2 weeks and with $100 in buyers’ price sensitivities can lead sellers to underprice their
4 weeks. goods and consequently forgo profits. Thus, the managerial
From the analysis above, we can also expect intriguing implication is that sellers should set higher prices than they
JE–SE preference reversals for options involving tradeoffs typically do for unfamiliar products (e.g., horse rides, a new

351
352 Hsee and Zhang

type of cell phone that transmits smell), but not for familiar positive correlations? One viable explanation, as mentioned
products (e.g., taxi, a conventional cell phone). earlier, is that higher income has led to improvement of inher-
ently evaluable variables, allowing individuals living in richer
regions or richer times to enjoy better inherently evaluable
Wealth and Happiness aspects of life, and because evaluation scales for these variables
A question that weighs on the minds of many is whether are not region- or time-specific, these individuals are abso-
increasing the wealth from one generation to the next makes lutely happier than those living in poorer regions or times.
the next generation happier. Numerous scholars have Another possible reason is that people in different countries
conducted important research on this question (e.g., Diener & may share some common standards of material well-being,
Biswas-Diener, 2002; Diener, Kahneman, Tov, & Arora, in such as having a television, having a car, and having one’s own
press; Diener, Ng, Harter, & Arora, in press; Diener, Sandvik, apartment in a city, and the standards, though socially deter-
Seidlitz, & Diener, 1993; Easterlin, 1974, 1995; Veenhoven, mined, may have remained relatively stable over the last few
1991) and have produced mixed findings. Some studies show decades. Thus, as the percentage of people owning these mate-
that increasing the wealth of all does not increase happiness rial goods increases, the average happiness also increases.
of all (e.g., Easterlin, 1974, 1995). Other studies suggest that However, we suspect that the standards for such learned-
raising wealth can increase reported subjective well-being and evaluable variables will sooner or later change as time evolves,
that life satisfaction in developed countries is on average higher whereas standards for inherently evaluable variables will not.
than that in less developed ones (e.g., Clark, Frijters, & Shields, One may also wonder if the gain in happiness brought about
2008; Diener, Kahneman, et al., in press; Diener, Ng, et al., in by improving inherently evaluable aspects of life will be
press; Diener et al., 1993; Kahneman, 2008; Stevenson & eventually erased by hedonic adaptation (e.g., Brickman &
Wolfers, 2008). Campbell, 1971; Brickman, Coates, & Janoff-Bulman, 1978;
According to our theory, a cross-generation increase in Diener, Lucas, & Scollon, 2006; Helson, 1964). We doubt it.
wealth may or may not lead to a cross-generation increase in Consider the findings of the temperature/jewelry study.
happiness depending on how the increased wealth is used. If Respondents in that study had presumably experienced their
it is used to improve inherently evaluable aspects of life, such room temperature many times and adapted to it. However, the
as sanitary systems, indoor temperature, childhood mortality respondents still showed a significant between-city effect of
rate, and physiological and psychological health, it will room temperature on their happiness.
increase happiness across time; otherwise, it may not. The find- One of the major puzzles in happiness research is determin-
ing of the temperature/jewelry study reviewed earlier (Hsee ing which variables are resistant to hedonic adaptation and
et al., 2009) is consistent with this proposition. In that study, which are prone to it. Our research can potentially shed light
each city can be compared to a separate generation, with richer on this question. We conjecture that the rate of hedonic adapta-
cities to richer generations. The finding suggests that although tion is related to the degree of inherent evaluability. Specifi-
both room temperature and jewelry value exert significant cally, inherently evaluable attributes are more resistant to
within-generation effects, only room temperature exerts a hedonic adaptation than inherently inevaluable attributes,
between-generation effect. because the former are associated with stable innate evaluation
One might wonder why we only recommend improving scales and the latter are not. Inherently evaluable attributes are
inherently evaluable aspects of life and not learned-evaluable like a foam ball. The foam ball has an inherent shape; although
aspects of life. The reason is that inherent evaluability (nature) the ball can be temporally squeezed by an external force, it will
and learned evaluability (knowledge) differ in a critical way. eventually return to its original shape. This analysis implies
Individuals’ reference systems and evaluation scales for inher- that inherent evaluability is a cause, not a result of (lack of)
ently evaluable variables (e.g., what temperature feels comfor- hedonic adaptation. That is, inherent evaluability can explain
table) are relatively stable across generations, whereas their why some variables are more resistant to hedonic adaption than
reference system and evaluation scale for learned-evaluable others, yet hedonic adaptation cannot explain why some vari-
variables are time-specific and will change as social norms ables are more inherently evaluable than others.
about these variables (e.g., which diamond size is considered
desirable) change. This distinction has not been made in the
previous research on evaluability (e.g., Hsee, 1996; Hsee
Summary
et al., 1999; Yeung & Soman, 2005, 2007). Consequently, Many diverse and important problems such as the relationship
improving inherently evaluable variables from one generation between demand and price, between wealth and happiness, and
to the next will make the next generation absolutely happier, between time duration and retrospective evaluation concern the
and improving learned-evaluable variables is merely a zero- same fundamental question: When are people value sensitive?
sum game and will only produce a treadmill effect. The GET advances our understanding of this fundamental
Recently, Diener, Kahneman, et al. (in press) and Diener, question. According to the theory, evaluability, hence value
Ng, et al. (in press) found positive correlations between income sensitivity, is high if evaluators have acquired relevant
and happiness both across countries and across generations in reference information through any of three main sources of refer-
the last few decades. How can our theory account for these ence information—mode, knowledge, and nature. Furthermore,

352
General Evaluability Theory 353

systematic overprediction of value sensitivity will arise if pre- Brickman, P., Coates, D., & Janoff-Bulman, R. (1978). Lottery
dictors are in a high-evaluability condition and predictees are winners and accident victims: Is happiness relative? Journal of
in a low-evaluability condition. Personality and Social Psychology, 36, 917–927.
In our opinion, the GET is parsimonious and generative. It is Chatterjee, S., Heath, T.B., & Min, J. (2009). The susceptibility of
parsimonious because it integrates several disparate lines of mental accounting principles to evaluation mode effects. Journal
research with only three basic factors. And it is generative of Behavioral Decision Making, 22, 120–137.
because the interplay between these factors yields a wide array Clark, A.E., Frijters, P., & Shields, M. (2008). Relative income, hap-
of testable hypotheses and provides a fertile ground for future piness and utility: An explanation for the Easterlin paradox and
research. other puzzles. Journal of Economic Literature, 46, 95–144.
Cryder, C.E., Mullen, E.E., & Loewenstein, G. (2008, November).
Notes Wanting versus choosing: A disconnect between what moves us
1. We assume in this article that evaluators at least know what the and what we prefer. Paper presented at the Society for Judgment
variable is about and which direction is more desirable. and Decision Making Preconference: Using Human Nature to
2. In Hsee and Zhang (2004), categorical differences are referred to as Improve Human Life, Chicago, IL.
qualitative differences, and incremental differences are referred to Desvousges, W.H., Johnson, F.R., Dunford, R.W., Boyle, S.P., &
as quantitative differences. Wilson, K.N. (1993). Measuring resource damages with contingent
3. There are four moderated mispredictions here, but only three mod- valuation. Tests of validity and reliability. In J.A. Hausman (Ed.),
erated effects in Proposition 2 because the order of factors (deter- Contingent valuation: A critical assessment (pp. 91–164).
mining which is the basic factor and which is the moderator) does Amsterdam: North-Holland.
not matter for moderated effects but does matter for moderated Diener, E., & Biswas-Diener, R. (2002). Will money increase subjec-
mispredictions. tive well-being? Social Indicators Research, 57, 119–169.
Diener, E., Kahneman, D., Tov, W., & Arora, R. (in press). Income’s
Acknowledgments association with judgments of life versus feelings. In E. Diener,
The authors thank the University of Chicago Center for Decision J. Helliwell, & D. Kahneman (Eds.), International differences in
Research and the Templeton Foundation for research support, and well-being New York: Oxford University Press.
thank Eva Buechel, Zach Burns, Paulina Celedon, Dan Gilbert, and Diener, E., Lucas, R.E., & Scollon, C.N. (2006). Beyond the hedonic
Lucy Shen for helpful suggestions on earlier drafts. treadmill: Revising the adaptation theory of well-being. American
Psychologist, 61, 305–314.
Declaration of Conflicting Interests Diener, E., Ng, W., Harter, J., & Arora, R. (in press). Wealth and hap-
The authors declared that they had no conflicts of interest with respect piness across the world: Material prosperity predicts life evalua-
to their authorship or the publication of this article. tion, while psychosocial prosperity predicts positive feeling.
Journal of Personality and Social Psychology.
References Diener, E., Sandvik, E., Seidlitz, L., & Diener, M. (1993). The rela-
Ariely, D., & Loewenstein, G. (2000). When does duration matter in tionship between income and subjective well-being: Relative or
judgment and decision making? Journal of Experimental Psychol- absolute? Social Indicators Research, 28, 195–223.
ogy: General, 129, 508–529. Dunn, E.W., Wilson, T.D., & Gilbert, D.T. (2003). Location, location,
Ariely, D., Loewenstein, G., & Prelec, D. (2003). ‘‘Coherent arbitrari- location: The misprediction of satisfaction in housing lotteries.
ness’’: Stable demand curves without stable preferences. Quarterly Personality and Social Psychology Bulletin, 29, 1421–1432.
Journal of Economics, 118, 73–105. Easterlin, R.A. (1974). Does economic growth improve the human
Bazerman, M.H., Loewenstein, G., & Blount White, S. (1992). Rever- lot? In P.A. David & M.W. Reder (Eds.), Nations and house-
sals of preference in allocation decisions: Judging an alternative holds in economic growth: Essays in honor of Moses Abramovitz
versus choosing among alternative. Administrative Science Quar- (pp. 89–125). New York: Academic Press.
terly, 37, 220–240. Easterlin, R.A. (1995). Will raising the incomes of all increase the
Bazerman, M.H., Schroth, H.A., Shah, P.P., Diekmann, K.A., & happiness of all? Journal of Economic Behavior and Organization,
Tenbrunsel, A.E. (1994). The inconsistent role of comparison oth- 27, 35–47.
ers and procedural job descriptions: Implications for job accep- Festinger, L. (1954). A theory of social comparison. Human Relations,
tance decisions. Organizational Behavior and Human Decision 7, 117–140.
Processes, 60, 326–352. Fredrick, S., Loewenstein, G., & O’Donoghue, T. (2002). Time
Bazerman, M.H., Tenbrunsel, A.E., & Wade-Benozi, K.A. (1998). discounting and time preference: A critical review. Journal of
Negotiation with yourself and losing: Understanding and manag- Economic Literature, 40, 351–401.
ing competing internal preferences. Academic Management Fredrickson, B.L., & Kahneman, D. (1993). Duration neglect in
Review, 23, 225–241. retrospective evaluations of affective episodes. Journal of Person-
Brickman, P., & Campbell, D.T. (1971). Hedonic relativism and ality and Social Psychology, 65, 45–55.
planning the good society. In M.H. Appley (Ed.), Adaptation Gilbert, D.T., Gill, M.J., & Wilson, T.D. (2002). The future is now:
level theory: A symposium (pp. 287–302). New York: Temporal correction in affective forecasting. Organizational
Academic Press. Behavior and Human Decision Processes, 88, 430–444.

353
354 Hsee and Zhang

Gilbert, D.T., Pinel, E.C., Wilson, T.D., Blumberg, S.J., & method. Organizational Behavior and Human Decision Processes,
Wheatley, T. (1998). Immune neglect: A source of durability bias 95, 20–39.
in affective forecasting. Journal of Personality and Social Psychol- Kogut, T., & Ritov, I. (2005). The singularity effect of identified vic-
ogy, 75, 617–638. tims in separate and joint evaluations. Organizational Behavior
Gonzalez, R., & Wu, G. (1999). On the shape of the probability and Human Decision Processes, 97, 106–116.
weighting function. Cognitive Psychology, 38, 129–166. Larrick, R.P. (2004). Debiasing. In: D.J. Koehler & N. Harvey
Gonzalez-Vallejo, C., & Moran, E. (2001). The evaluability hypoth- (Eds.), Blackwell handbook of judgment and decision making
esis revisited: Joint and separate evaluation preference reversal (pp. 316–337). Oxford, United Kingdom: Blackwell.
as a function of attribute importance. Organizational Behavior and List, J. (2002). Preference reversals of a different kind: The ‘‘more
Human Decision Processes, 86, 216–233. is less’’ phenomenon. American Economic Review, 92,
Helson, H. (1964). Adaptation-level theory. New York: Harper & Row. 1636–1643.
Hsee, C.K. (1996). The evaluability hypothesis: An explanation for Loewenstein, G., O’Donoghue, T., & Rabin, M. (2003). Projection
preference reversals between joint and separate evaluations of bias in predicting future utility. Quarterly Journal of Economics,
alternatives. Organizational Behavior and Human Decision Pro- 118, 1209–1248.
cesses, 67, 247–257. Loewenthal, D. (1993). Reversals of preference in candidate choice.
Hsee, C.K. (1998). Less is better: When low-value options are valued Unpublished manuscript, Carnegie-Mellon University, Pittsburgh,
more highly than high-value options. Journal of Behavioral Deci- PA.
sion Making, 11, 107–121. Moore, D.A. (1999). Order effects in preference judgments: Evi-
Hsee, C.K., Blount, S., Loewenstein, G., & Bazerman, M. (1999). dence for context dependence in the generation of preferences.
Preference reversals between joint and separate evaluations of Organizational Behavior and Human Decision Processes, 78,
options: A review and theoretical analysis. Psychological Bulletin, 146–165.
125, 576–590. Morewedge, C.K., Kassam, K.S., Hsee, C.K., & Caruso, E.M. (2009).
Hsee, C.K., & Rottenstreich, Y. (2004). Music, pandas, and muggers: Duration sensitivity depends on stimulus familiarity. Journal of
On the affective psychology of value. Journal of Experimental Experimental Psychology: General, 138, 177–186.
Psychology: General, 133, 23–30. Paharia, N., Kassam, K.S., Greene, J.D., & Bazerman, M.H. (2009).
Hsee, C.K., & Shen, L. (2009). Taxi versus horse: What affects con- Dirty work, clean hands: The moral psychology of indirect agency.
sumer price sensitivity and when price-setters mispredict it? Organizational Behavior and Human Decision Processes, 109,
Unpublished manuscript. 134–141.
Hsee, C.K., Yang, Y., Li, N., & Shen, L. (2009). Wealth, warmth, and Read, D. (2004). Intertemporal choice. In: D.J. Koehler & N. Harvey
well-being: Whether happiness is relative or absolute depends on (Eds.), Blackwell handbook of judgment and decision making (pp.
whether it is about money, acquisition, or consumption. Journal 424–443). Oxford, United Kingdom: Blackwell.
of Marketing Research, 46, 396–409. Read, D., Frederick, S., Orsel, B., & Rahman, J. (2005). Four scores
Hsee, C.K., & Zhang, J. (2004). Distinction bias: Misprediction and and seven years from now: The date/delay effect in temporal dis-
mischoice due to joint evaluation. Journal of Personality and counting. Management Science, 51, 1326–1335.
Social Psychology, 86, 680–695. Rottenstreich, Y., & Hsee, C.K. (2001). Money, kisses, and electronic
Irwin, J.R., Slovic, P., Lichenstein, S., & McClelland, G. (1993). Pre- shocks: On the affective psychology of risk. Psychological Sci-
ference reversals and the measurement of environmental values. ence, 12, 185–190.
Journal of Risk and Uncertainty, 6, 5–18. Schkade, D., & Kahneman, D. (1998). Does living in California make
Kahneman, D. (2008). The sad tale of the aspiration treadmill. people happy? A focusing illusion in judgments of life satisfaction.
Retrieved May 21, 2009, from http://www.edge.org/q2008/q08_ Psychological Science, 9, 340–346.
17.html#kahneman Schmeltzer, C., Caverni, J., & Warglien, M. (2004). How does prefer-
Kahneman, D., Fredrickson, B.L., Schreiber, C.A., & ence reversal appear and disappear? Effects of the evaluation
Redelmeier, D.A. (1993). When more pain is preferred to less: mode. Journal of Behavioral Decision Making, 17, 395–408.
Adding a better end. Psychological Science, 4, 401–405. Soman, D., Ainslie, G., Frederick, S., Xiuping, L., Moreau, P.,
Kahneman, D., & Ritov, I. (1994). Determinants of stated willingness Mitchell, A., et al. (2005). (2005). The psychology of intertemporal
to pay for public goods: A study in the headline method. Journal of discounting: Why are distant events valued more than proximal
Risk and Uncertainty, 9, 5–38. events? Marketing Letters, 347–360.
Kahneman, D., & Snell, J. (1992). Predicting a changing taste: Do Stevenson, B., & Wolfers, J. (2008). Economic growth and subjective
people know what they will like? Journal of Behavioral Decision well-being: Reassessing the Easterlin paradox. Bookings Papers on
Making, 5, 187–200. Economic Activity, 1–87.
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of Tehila, K., & Ritov, I. (2005). The singularity effect of identified vic-
decision under risk. Econometrica, 47, 263–292. tims in separate and joint evaluations. Organizational Behavior
Kluger, A.N., Stephan, E., Ganzach, Y., & Hershkovitz, M. (2004). and Human Decision Processes, 97, 106–116.
The effect of regulatory focus on the shape of probability- Van Boven, L., Loewenstein, G., & Dunning, D. (2005). The illusion
weighting function: Evidence from a cross-modality matching of courage in social predictions: Underestimating the impact of

354
General Evaluability Theory 355

fear of embarrassment on other people. Organizational Behavior Yeung, C., & Soman, D. (2005). Attribute evaluability and the range
and Human Decision Processes, 96, 130–141. effect. Journal of Consumer Research, 32, 363–369.
Veenhoven, R. (1991). Is happiness relative? Social Indicators Yeung, C., & Soman, D. (2007). The duration heuristic. Journal of
Research, 24, 1–34. Consumer Research, 34, 315–326.
Willemsen, M.C., & Keren, G. (2004). The role of negative features in Zhang, J., & Hsee, C.K. (2009). Distinction bias: The moderating role
joint and separate evaluation. Journal of Behavioral Decision Mak- of attribute evaluability. Unpublished manuscript.
ing, 17, 313–329. Zikmund-Fisher, B.J., Fagerlin, A., & Ubel, P.A. (2004). ‘‘Is 28%
Wilson, T.D., & Gilbert, D. (2003). Affective forecasting. Advances in good or bad?’’ Evaluability and preference reversals in health care
Experimental Social Psychology, 35, 345–411. decisions. Medical Decision Making, 24, 142–148.

355