The Hunt for Better Decisions

Making better decisions can be seen from several perspectives. One draws a line between “good” decisions and “bad” decisions. Such a point of view is traditionally linked to Decision Theory, the discipline that provides us with the basic normative principles in choice, hence allowing for good decisions.

A second perspective, based on observed behavior in the laboratory and in the field, focuses on the study of “optical illusions” of the mind (cognitive distortions) that could result in decision making biases. Decision Analysis combines both normative and descriptive aspects of Decision Theory to help take decisions in real choice situations, and hence provide prescriptions. 

Professor Itzhak Gilboa from HEC Paris proposes two contributions. The first explains how analyzing decisions is a part of our humanity, and to what extent Decision Theory could help decision makers. In a second contribution, Itzhak Gilboa offers an account of how economic models can be useful despite the fact that they make unrealistic assumptions, and, in particular, analyze the behavior of rational decision makers while psychological studies show that normal people systematically depart from rationality.

Then, CNRS Research Professor Brian Hill tackles the crucial issue of how decision makers should decide in the face of radical uncertainties such as those regarding climate change or genetically modified organisms. In particular, he underlines the importance of confidence in rational decision. 

In the fourth contribution, CNRS Research Professor Emmanuel Kemel shares the results of a series of laboratory experiments which study behavioral differences in decision making when consequences are measured in units of time rather than in money. The author explains why people manage time differently to managing money.

The newsletter concludes with the interview of CNRS Research Professor Mohammed Abdellaoui who shares his experience as a researcher in Decision Theory and Behavioral Decision Making over the two last decades.

Bespectacled and affable, Mohammed Abdellaoui does not cut a figure of a hard-nosed theoretician at the cutting-edge of one of the West’s oft-overlooked sciences. Since 1992, the researcher has been studying and modeling our perceptions of risk and time when it comes to making decisions. In this, he is following footsteps that go back to the philosopher and mathematician Blaise Pascal. In the 17th century the Frenchman made a wager on people’s belief in God. This is seen as the first use of decision theory, laying the foundations for schools of thought such as existentialism, voluntarism and pragmatism. 

330 years later, Abdellaoui, freshly out of the University of Aix-Marseille III with a Ph.D. in mathematical economics and econometrics under his arm, joined the CNRS to study behavioral aspects of decision making. Reclining in his modest campus office in HEC’s W1 wing, he explains why: “In 1992, there was almost no experimental investigations about individual decision-making here in France. I was one of the first to run experimental investigations on individual behavior to test new decision theories.”

But how would Abdellaoui define this discipline which explores the reasoning behind a person’s choices? “To begin with, you have to understand that there are a number of overlaps: in Decision Theory, the object of study is the human being, so you have to deal with his/her psychology, preferences, beliefs. Traditionally, the discipline uses mathematical tools to model individual behavior. But the modern study of individual behavior under uncertainty needs more than that. We therefore combine maths, statistics, econometrics and microeconomics with psychology and philosophy.” Sounds complicated. “Yes,” he replies matter-of-factly, “you have to exhibit a lot of skills if you want to enjoy a career in this field.” 

Abdellaoui is interested in the study of decision making from two points of view. The first, descriptive, regards how people make their decisions. The second point of view is normative. It focuses on how to make the ‘best decisions’ while accounting for the decision maker’ beliefs and tastes. “Normativity is what distinguishes decision theory from behavioral psychology. Psychologists are not interested in normative rules involved in decision-making. They take behavior as it is, observing without judging. But economists and decision theorists must isolate what is ‘good’ and ‘bad’ in decision-making.”

The researcher is also editor-in-chief of Theory and Decision, an international multidisciplinary journal focusing on advances in Decision Science. He has built models which attempt to show us how to take decisions in the face of uncertainty and how to elaborate new rules of consistency and rationality. These aim at allowing business executives, policy-makers and individuals to make “good decisions” based on normative choice theories. He uses decision-aiding tools such as decisions trees, influences diagrams and Bayesian Networks (named after an 18th century statistician, this translates the state of the world into degrees of belief). Some of these are drawn on his office blackboard presenting a bewildering tapestry of algebraic formulae and equations.

“This science has evolved in a radical way in the past 30 years. We had to catch up with Anglo-Saxon researchers who pioneered new concepts of decision analysis in the Fifties and Sixties.” Abdellaoui picks out the remarkable career of Harvard University academic Howard Raiffa who passed away last July aged 92. Raiffa’s works on betting on horses led to US authorities using his Bayesian methods extensively, including the search for a missing US Air Force hydrogen bomb which disappeared off of Spain in 1966. “By then, American research had become predominant. Many big US companies and governmental agencies like NASA appealed to decision analysts and academic researchers to help them in the decision-making process. There is an American tradition of delegating when it comes to situations that are out of reach of a group’s expertise. This explains why applications of decision theories are most developed in the US.”

But, over the past decades, French academia - with HEC Paris at the forefront - have invested heavily in the field. And it shows: researchers in France followed the footsteps of the 1988 Nobel laureate Maurice Allais with pioneering work of their own, eagerly transmitted in university circles. “We provide students (Grande Ecole and MBA) with modern tools for modeling and solving complex decision-making challenges. We have become more and more international and connected to global research networks. As a result, progress in research on behavioral economics and decision-making in France is huge.” Abdellaoui was attracted to the dynamism of GREGHEC and joined the team at Jouy-en-Josas in 2007. “I was interested in how the business community makes daily decisions. At HEC, we compare observed behavior to the normative model and check if this is consistent or not. Here, I found the right environment for my field: international, a team of young researchers experimenting in real settings, and established figures like Itzhak Gilboa leading the group.” 

The diversity of axes of inquiry have also coalesced. “Beforehand, some researchers focused on decision-making under uncertainty, others studied it over time. Because uncertainty and time are intertwined, we are more interested in investigating individual behavior involving both time and uncertainty nowadays. We’ve therefore elaborated more complex models.” The key remains international exchanges, however. HEC Paris’ small team of researchers has acquired an unparalleled reputation for normative models and applications that work on ambiguity in decision-making. “Our Ph.D. students collaborate with the likes of Peter Klibanoff at Northwestern University, Manel Baucells at University of Virginia, Peter Wakker at Erasmus University. We also have a joint program with the University of Berlin, in India, where we’re studying decision-making over time in agriculture. A better understanding of the psychology of risk and time in such settings would allow efficiently using funds from international organizations. The results could go beyond India. In Africa, for example, international NGOs could invest more intelligently instead of making people dependent on (charitable) assistance.” 

Hard decision-making applications infuse a surprising diversity of fields. Abdellaoui enumerates a few, first on the microeconomic level: “You have the psychology of time and risk playing on people with AIDS in Africa. Without prior access by victims to appropriate health care programs, funding provided by organizations could not result in the improvement of their situation. Equally, the production of low-cost goods are subject to decision theories: if you know better temporal and risk preferences of decision-makers, you can adapt products to fit them.”

“Decision-making models also come into play for major economic players, like the transport giants SNCF and RATP, or aviation authorities who have to decide on the feasibility of building another airport in Notre-Dame-des-Landes or outside London. In these cases, you can make trade-offs. Decision-making is closely related to trade-offs. Take, for example, the increase in construction cost which is traded off with the drop in environmental concerns. An airport has multiple criteria that is either substitutable or complementary: access, hotel availability, cost, environmental damage, etc. Such criteria is relevant and help the authorities make a decision in difficult circumstances, using normative models.”

To vehicle such theories, decision scientists have turned to the language of mathematics and mathematical models. “It simplifies decision problem keeping important aspects within the picture,” explains Abdellaoui. He points to an equation on the blackboard. For the layman, it is indecipherable: W of p exponential of +/- RT times v of c. “This represents the value of a risky prospect today” the professor explains kindly. “And this is the value a year later, which is possibly less. We can convert our optimism in the future with this sign. And the ingredient of the model could explain that optimism.”

The researcher is optimistic about the future of decision sciences and long-term international collaborations. “Research is now global, we work alongside people with different expertise and skills, with easy access to data. We have to gather all this to create viable collaborations. So international cooperation is a must.” But the professor who also conducts research on the campus of Al Akhawayn, in the beautiful town of Ifrane, Morocco, remains guarded about isolationist calls emanating from Europe and the USA. “It’s a complex question,” he mutters, “Isolationism could negatively impact academic exchanges. Additionally, it could distort the rules of the game regarding international trade. For instance, if the US refrains from signing international agreements to reduce fossil fuels consumption, that would expose businesses to additional risks which have to be factored in.”

Meanwhile, Mohammed Abdellaoui prides himself in working in an environment where researchers and students are publishing major works on decision theory and applications in leading international reviews. He also insists on how his research could contribute to avoid cognitive distortions and biases in individual decision-making. “Students are provided with a catalogue of stylized facts in individual decision making. This shows them how to circumvent biases regarding preferences and beliefs.” An objective which is likely to spill into the business world when graduates embark on their own careers.

The decision sciences have made great improvements over the past decades in our understanding of how people make decisions involving uncertainty. By contrast, little progress has been made on the normative question: how should people choose in the face of uncertainty? And yet, the established paradigm around this question – developed by philosophers, economists and statisticians in the middle of the last century, and often called Bayesianism – seems incapable of coping with the toughest uncertainties facing us today.

An on-going project, led by Brian Hill at the HEC Paris-CNRS Decision Sciences laboratory and partly funded by the French Agence Nationale de Recherche (projet DUSUCA), deals precisely with this normative question. To do so, it has been involved in developing, defending and promoting a new model for decision making in the face of uncertainty, based on a notion that has been largely neglected to date: confidence.

The role of confidence in decision-making

Uncertainty denotes a lack of knowledge or sufficiently strong belief. Any incorporation of uncertainty into a decision involves a consideration of what the decision maker can justifiably believe. Beyond the fact of holding a certain belief, one can be more or less confident in it. People seem to be more confident in certain beliefs than others; compare, for example, investors’ confidence in their beliefs about the performance of a certain stock in a familiar market with how sure they are about their opinions concerning a market with which they have no experience. The level of confidence in beliefs seems to play a significant role in decisions.

Ideally, one would like environmental policy decisions to be based on judgements about which we are fairly confident (in view of the evidence accumulated by decades of research, for instance), rather than on mere hunches (for instance, the uncorroborated hunch that the issue of climate change is Chinese conspiracy).

In a series of papers, Brian Hill develops a model of decision making that incorporates the degree of confidence in beliefs. It does so according to the maxim that the more important a decision is, the more confidence is required of a belief for it to be used in making that decision. In other words, hunches can be used for unimportant decisions, but when the stakes are high, one should only rely on beliefs in which one has considerable confidence.

I see the climate change example as a strong case study for the model that we have developed.

Confidence, and how you should decide in the face of uncertainty

Economists and philosophers have evoked various considerations to evaluate the normative credentials of  decision-making procedures. Many are based on abstract mathematical results that reveal whether a prima facie reasonable procedure leads decision makers to make embarrassing decisions in certain situations.

Several published and some more recent papers, developing intra alia results of this sort, make the case that Brian Hill’s theory of decision making and confidence provides a reasonable account of how one should go about deciding. These investigations suggest that his proposed decision-making model and its incorporation of confidence in beliefs has significant advantages, as a normative account, over other existing proposals in the literature.

Climate uncertainties and climate decisions

Recent and current work has focused on connecting Brian Hill’s theory of decision making with real-life decisions in the face of severe uncertainty. Perhaps one of the most striking examples is climate policy. The Intergovernmental Panel on Climate Change (IPCC) periodically summarizes the present state of scientific knowledge about climate change and its impacts. The main goal of this exercise is to inform environmental policy making. However, the IPCC uses a specific technical language for conveying uncertainty, which people do not seem to know how to connect to existing decision-making processes. Certainly, it is unclear that policy makers are making full and proper use of the information provided to them in the IPCC’s reports.

According to Brian Hill, however, his decision-making model is particularly well-suited to harnessing the information supplied in these reports. “I see the climate change example as a strong case study for the model that we have developed,” he explains. “Our model can use all of the information provided in the IPCC’s reports without needing any more. This is not the case for other existing models: either they cannot process parts of the information the IPCC provides, and so have to ignore them, or they require something that the IPCC does not supply. To be more specific, some approaches reserve no role for confidence in beliefs, and so are too simple to take into account the confidence information provided by the IPCC. On the other hand, models that are rich enough to represent a confidence dimension assume that it takes precise numerical values, without which the models cannot be applied.

However, the IPCC provides no such numerical values, instead using a qualitative 5-point evaluation of confidence levels, ranging from very low to very high. Indeed, the philosophy behind our model is in line with the IPCC’s thinking on this issue: there seems to be no non-arbitrary way to characterise with a precise number the confidence level that the current state of scientific research warrants in a given finding when it comes to climate change – especially when this research involves a range of scientific studies, in different disciplines, with different data, different methodologies, and perhaps even different conclusions.” Joint work with colleagues at the LSE suggests that Brian Hill’s model is the first developed that can deal with the kinds of important decision-making inputs provided by the IPCC reports. In this way, it also vindicates the IPCC’s technical uncertainty language as normatively justifiable, by showing that it fits well into a reasonable decision-making procedure. 

Brian Hill concludes that the model’s appropriateness in regards to climate change suggests that it could be applied to other decisions that must be made in the face of uncertainty. “The IPCC represents one of humanity’s biggest efforts to document the level of  knowledge and uncertainty on an issue, and our model provides decision makers with an ideal way of harnessing the incredibly complex and rich information they supply. This suggests that the approach can make valuable contributions to other decisions under severe uncertainty, across a wide range of situations and sectors.”

Researchers in other social sciences, as well as laypeople, often question why economic models analyze the behavior of “homo economicus”, when psychological research shows it to be an unrealistic model.  Why are economists relatively unperturbed by the violations of their models’ assumptions in experiments?  How can oversimplified and “toy” models be of use in making predictions?  These methodological problems have been discussed since the 1950s, most famously by Milton Friedman, and have received more attention in recent decades. 

Approaching models as analogies that can inform predictions

Akerlof’s famous “Lemon Market” is a simple economic model that can be easily explained to students in a single class. Some question how so simple a model could possibly merit the Nobel Prize. Others demand what economists could possibly learn from such an idealized and unrealistic model. One explanation in defense of the value of this model – as well as the value of this type of model in economic research more broadly – is that the standard notion of scientific prediction based on general, unrefuted theories is only one way to make predictions; another way involves analogies.

This argument parallels the distinction between rule-based and case-based reasoning in psychology and artificial intelligence, whereby people think in terms of generalizations (“rules” and “theories”) but also in terms of analogies to similar cases.  Since such distinctions can be found between statistical techniques for generating predictions, it stands to reason that social scientists might also employ both ways of thinking. Viewed thusly, a formal mathematical model is not to be read as a general statement, “whenever… we will observe…” but only as a report of a (theoretical) case: “in this hypothetical story analyzed on my whiteboard, I found that…” Similarly, an experiment run in a lab is also a case, but one that reports a real occurrence. 

Taking both theoretical models and lab experiments into account

As individual cases do not involve universal quantifiers, they cannot contradict each other.  In particular, when faced with a real-life economic problem, the economist has to judge whether it is more similar to the theoretical case analyzed on the whiteboard or to the experiment run in the lab.  For example, consider the “ultimatum game”, in which Player I offers Player II a split of, say 100 euros, and Player II can accept or reject.  If she rejects, both get nothing.  Assuming the payoffs are only monetary (so that pride, equality, altruism, and other such considerations play no role), the theoretical prediction is that Player I will make an offer of 1 euro, and that Player II will accept it.  And yet, experiments show that this is often not the case.

People in the role of Player II often reject low offers, (perhaps because of pride) and in the role of Player I they often make much more generous offers (perhaps because they anticipate Player II’s response). If conceptualized as a general statement about such economic interactions, then the latter would seem to contradict the theoretical prediction. If the theoretical model and the lab experiment are conceptualized as individual cases, however, then it becomes possible to take both into account.  If we are asked to make a prediction about a real-life event, we could ask ourselves: is the event more similar to the theoretical model or the lab experiment?  Sometimes, the answer may be the theoretical model.  For example, if the payoffs are in millions of euros, we may predict that Player II will accept an offer of 1M euros, despite the fact that Player I will get 99 times more. 

Economists should be judged for their contribution to society by preventing silly mistakes or wrong-headed decisions.

Why economists do not worry about “refutation” 

In other words, we can better understand why economists are not so worried about “refutations” of their models, when the latter viewed as analogies, rather than as general theories. This also explains why economists prefer simple models: the simpler the model, the easier it is to see analogies.  If economic models are approached as analogies, does it make economics a refutable science?  Well, it depends. If the models are also equipped with “user manuals” that explain how to estimate similarity and what exactly to do with multiple “cases”, the answer is yes.  Different cases can be aggregated to generate predictions that can be confronted with the data. It should be recognized, however, that economic models often do not come with these “instructions”, leaving much room for subjective judgment in determining a model’s predictions, and thereby in the way it is tested. 

Reconsidering the role of economics 

A second way to explain economic models – and of economic analysis in general – is to consider economics as a field of criticism, rather than a predictive science. Viewed thusly, an economist should not be asked: “What will happen now?” or even “What will happen if…?” Rather, her job is to follow the public debate, and, if she finds a flaw in an argument, to point it out.  Perhaps a politician or a journalist suggests a line of reasoning that is logically flawed.  For instance, assuming that raising the tax rate will necessarily increase tax revenue, without taking into account equilibrium considerations.  In such a case, the economist’s job is to point out the problem in the reasoning, without necessarily offering alternative predictions.  According to this view, economists should be judged for their contribution to society by preventing silly mistakes or wrong-headed decisions.  In a sense, their social role as critics can be likened to those of the historians: the latter would typically say that it is not their job to make predictions, and yet we all think that studying history is extremely useful. 

Importantly, neither of these explanations – 1) of economic models as analogies rather than general theories and 2) of economics as a field of criticism rather than a predictive science – attempts to address how successful economics is or is not. In both papers, formal models are used to support informal arguments. Some results are proven and explain why economists face certain problems (in a formal sense) as well as also partly explaining why economists tend to favor seemingly over-simplified models. Overall, the models and the explanations that they provide make it easier to see analogies between different cases (at the expense of accuracy of description), as well as highlight the erroneous step in a proposed reasoning.

Most empirical studies in decision science focus on choices involving money. In contrast, important everyday decisions can also involve non-monetary attributes, such as time. Examples include choosing the best route when traveling, deciding whether to work or relax during a weekend, or expediting or postponing a time-consuming task. Classical economic models usually convert consequences measured in time units into their monetary equivalents, hence assuming that “time is money”. The assumption that people manage their time like their money has however not yet been investigated empirically within the framework of rational choice models. In a series of experiments, Mohammed Abdellaoui, Emmanuel Kemel and Cedric Gutierrez set out to do exactly that, observing how people make decisions involving gains and losses of time and how these choices differ from those involving money. Their results reveal that, at least in the laboratory, subjects take radically different decisions when time is involved as compared to money.

The difficulty of evaluating time

Despite the old saying that “time is money,” several important differences distinguish the two attributes. The main one is probably fungibility: unlike money, time cannot be stored nor saved, and the value of time is highly context dependent. While a loss of money can be compensated by a gain of money, the same is often not true for time. For example, during a leisure week-end in Venice, an unexpectedly high taxi bill has a different impact than losing several hours stuck in a traffic jam. While the loss of money can be compensated later, time lost in Venice cannot be compensated by extra free time when back home. Furthermore, contrarily to money, the value of time is not necessarily monotonic. Having more money is always better, but the pleasure of time dedicated to an enjoyable activity can decrease at some point. Enjoying a 2 hours movie does not mean that you would have preferred it to last 3 hours. Given these characteristics, the perception of time can give rise to paradoxical behavior. Past research has highlighted that people tend to systematically under-estimate the duration of past episodes (duration neglect), or the time needed to complete a given task (planning fallacy). Misperception of time can also lead to inconsistent behavior when people and organizations decide about the future (dynamic inconsistency). 

Several aspects of decisions are systematically different between time and money. One of them regards risk-taking.

Designing lab experiments with standardized definitions of time 

In a recent series of laboratory experiments, Mohammed Abdellaoui, Emmanuel Kemel and Cedric Gutierrez studied how people make choices when possible consequences are expressed in terms of gains and losses of time. Since the value of time can vary largely with the context, one of the challenges of the project was to design a standardized definition of time. This was achieved by the construction of concrete scenarios. In a first experiment, the time considered was the number of minutes that subjects spent in the experiment. Losses and gains of time were defined as the possibility to respectively increase or decrease the duration of the experiment (up to 1 hour) so that subjects could have to spend more or less time in the laboratory. Time thus referred to the same context for all subjects and the gains and losses of time could be implemented for real (i.e. subjects could really leave the experiment earlier than expected, or spend more time than expected on the experiment.) In another experiment, the scenario consisted of a job contract of two sessions of 4 hours each, at a sooner and a later period, for a given salary. Choices involved variations of working time. For instance, the experimenters observed whether subjects preferred to enjoy a 1-hour reduction of working time at a sooner or later period. The experiments also involved similar choices with monetary consequences that were used as a baseline treatment (or benchmark).

Observing how people decide about time versus money

Results show that several aspects of decisions are systematically different between time and money. One of them regards risk-taking. Subjects take fewer risks when deciding about losses of time than when deciding about losses of money. This may be because losses of time are harder to compensate than losses of money. Another striking difference regards how people perceive futures losses of time as compared to future losses of money. For money, rationality recommends delaying losses as far as possible in the future. In regards to time however, most subjects are indifferent to postponing time losses. A sizable share of subjects even exhibits a preference for expediting such losses. While it stands in opposition to financial rationality, such behavior is consistent with the popular wisdom: “never put off till tomorrow what you can do today.” As for money, subjects take different decisions whether consequences are framed in terms of gains or losses. This so-called “framing effect” has been largely documented for monetary choices, and is often used in marketing or public policy as a way to influence decisions. The researchers observed that asymmetries between gains and losses at the origin of the framing effect are more pronounced for time than for money. This suggests that framing can be an efficient tool to influence the way people manage their time. 

The results offer a better understanding of how people manage their time. In particular, they highlight particular choice patterns  that can be specific to such decisions. For instance, this research shows a large heterogeneity in choices involving time, suggesting that managers can expect a large diversity of behavior in the way people organize their time. It also shows that people are more vulnerable to decision biases, such as procrastination, when choices involve time. These findings can be used to develop management strategies that aim at improving time management in decision-making situations.

Analyzing decisions is human

Decision making is, indeed, associated with animals more than with plants, and with higher life forms more than with lower ones. Surely, plants may exhibit phenomena that can be described as “deciding” to do something. A plant that turns its branches to face the sun can be viewed as making a decision, and even as choosing the branches’ angles optimally. Yet, this description would be more of a metaphor, a personification of the plant, than an actual account of decision making. A detailed process of encountering a problem, pondering it, and deciding what to do is more typical of higher life forms with relatively large brains than of lower ones. A tiger who decides if and when to start chasing its prey, a crow who decides how to break a nut’s shell, an alpha-male ape who decides whom to grant privileges to in its clan – all seem to go through stages of thought, starting with noticing the decision problem, going through deliberation, all the way to making a decision. At times, animals can even have emotional reactions not only to the decision’s outcome, but also for having made it, as in the case of a dog who seems to be ashamed, guilty, or embarrassed about a decision it should not have made.

Thus, the human race is clearly not the only species who makes decisions. By contrast, analyzing decisions does seem to be much more typical of humans than of animals.[1] Decision analysis has many manifestations, occurring prior to, during, and following a decision. In particular, a decision that we think of as conscious requires that one imagine the different states of affairs that can be brought about by different possible choices. The ability to depict these should not be taken for granted. In particular, it requires that one be able to suspend some knowledge one might have about oneself. Even if a person’s choice is easily predictable from choices she has made in similar past cases, rational decision making still requires that this prediction be put on hold, to allow her to contemplate her habitual choices and the possibility of changing them. This mental exercise of suspending self-knowledge is probably essential to the psychological phenomenon of free will: we would probably not have the internal experience of free will if we could not entertain several images of the future concurrently. Similarly, learning from past decisions requires that one be able to reason about counterfactuals, imagining what would have transpired had one made different choices. The psychological phenomenon of thinking in causal terms is thus tightly related to our ability to analyze our decisions.

Decision theory was developed with an emphasis on the human

If decision analysis is a hallmark of humanity, we’d expect it to be present in any account of humanity, including the most ancient ones. Indeed, recorded history is replete with the analysis of decisions: recommendations for optimal decisions; reasoning that leads to decisions; contemplation of alternative decisions, and so forth. Moreover, some of the insights modern theory can provide into decision making have a rather respectable history. Ask an investment analyst how for advice, and they may suggest diversification, namely, splitting the investment among several assets, so as to minimize risk. The preference for lower risk and this idea of diversification already appear in Genesis, where Jacob is about to face his brother Esau, and, fearing the encounter, splits his camp in two, so that one half might be saved should Esau smite the other. Similarly, dealing with self-discipline problems, we might recommend various ways of self-commitment. But in so doing we will be following Ulysses, who tied himself to the mast of his boat so as to enjoy the Sirens’ singing but not be tempted to swim to the dangerous rocks. 

And yet, decision theory as a field of inquiry hasn’t been developed before the 17th century. It wasn’t enough to be human to develop such a theory; it was necessary that humanity be celebrated. This process, which in western culture (re-)started with the Renaissance and was advanced by the Enlightenment, put the human at center stage. This had many manifestations, some of which resurrected the Classics, and some was truly novel. It included dramatic changes in music, architecture, and the arts; the development of modern science; and also the invention of some aspects of social science. It took awarding humans a central status to make their decisions an appropriate object of study. 

If we had to select one point in which decision theory started, it would probably be Pascal’s wager. It is perhaps not too surprising that one of the people most closely associated with the invention of probability theory and mathematical expectation is also the one who first thought about decision making in an orderly way. The wager is considered a very modern, post-Renaissance type of question: Pascal doesn’t attempt to prove that God exists; rather, that one would do wisely to become a believer (at least in due course). The locus of analysis isn’t the universe, in which God is, or is not. Rather, it is the human mind, wherein a belief in God may reside. Moreover, while Pascal is careful not to assume that one can simply choose one’s beliefs, he does frame the question as one of choice.

In his famous analysis of the wager, Pascal lays the foundations for several important ideas of decision theory. He informally describes the decision matrix, drawing a distinction between acts, over which one has control, and states or events, over which one doesn’t. He describes the notion of a dominant strategy. He then proceeds to refer to subjective probabilities – introducing the idea that the machinery of probability theory, developed to deal with objectively given probabilities as in chance games, can be used to organize one’s beliefs and decision making. Pascal then introduces his main argument, which is basically that of expected utility maximization, and concludes with the adaptation of his argument to the case of unknown probabilities.

It is no coincidence that decision theory wasn’t developed in medieval times. There was little room for such a theory when center stage was occupied by deities, and the main ways to deal with an uncertain future were prayer and penitence. Only when humans turned the spotlights to themselves did decision theory come into being.

Theory cannot replace human decisions; it should complement them

With the rise of mathematical economics, operations research, and related fields, decision theory and game theory were also developed in the mid-20th century. The foundations laid by luminaries such as John von Neumann and Oskar Morgenstern; Frank Ramsey, Bruno de Finetti, and Leonard Savage; John Nash, Kenneth Arrow, Gerard Debreu; Lloyd Shapley and Robert Aumann gave much hope for the ability to understand decision making in a mathematical way, providing both predictions and recommendations. However, this promise was not fully fulfilled. Despite some huge advances in problems such as operations research, closer to computer science and engineering, much seemed to be lacking when it came to everyday life problems, ranging from macroeconomic and political systems to individual decisions involving personal values, unique events, and so forth. At times, it appeared that the problem is one of information, where notions such as utility and probability are hard to assess, whether for predicting people’s choices or for making recommendations for decision makers. At other times, it seemed that more is missing. Indeed, following the pioneering works of Daniel Kahneman and Amos Tversky, psychologists showed that practically all the assumptions of rational choice theory can be refuted in carefully designed experiments. These studies clearly showed that classical decision theory was lacking as a description of human behavior.

Worse still, they also raised doubts about its appropriateness as a normative standard. In many of these cases not only did the theory fail to be a good description of how people make decisions; it was also considered to be impractical or unconvincing.

The disenchantment with decision theory occurs when modernism is on the defense. For various reasons, authority is being questioned in general, and scientific authority is no exception. In such an atmosphere there is a risk that decision theory will be perceived as a total failure, to be replaced by sheer intuition. There are people who believe that intuition leads us astray, and that this cannot be fixed; and there are people who believe that intuitive decision making is very effective when things matter, and when people are in their natural environment. Common to both is the conclusion that there is no room for the development or study of the theory, as it is anyway useless. 

I hold that this would be a serious mistake. True, only in certain domains can decision theory fully replace intuition and provide the “best” recommendation. But in most if not all situations it can still be useful as a set of guidelines, as an aid to decision making. If we adopt a perhaps more realistic, more mature view of the theory, we may find that it can help us make better decisions – in our own eyes – also when it cannot mathematically model the problem with precision. In some cases, as in finding the shortest path between two points on Google Maps, decision theory can be trusted to replace intuition. But in many other cases – as in deciding what career to choose, where to invest one’s money, and so on – it can help us avoid mistakes without necessarily “computing” the correct answer.

Analyzing our decisions is an essential part of our humanity. It would be a pity to lose it. The analysis should take the human into account, which means that rational thought should be in a dialog with intuition and emotions. It is perhaps this dialog which is a quintessential ingredient of being human.

1 - This may partly be an artifact of our language: for all we know, it is possible that apes and dolphins analyze decisions as well, only lack the means to convey this analysis to us. For the sake of the present discussion, we will view decision analysis as a defining feature of humankind, and allow for the possibility that this is another way in which apes can be human.