User:Gilbert Peffer/Draft article: Modeling of emotions in finance

Overview
The dominant paradigm in finance is that of perfect rationality and efficient markets, where investors and traders filter and process information with a cold head, and where small price discrepancies are quickly identified and exploited by sophisticated arbitrageurs. Nobody negates the existence of emotions in financial decision making, but financial economics subsumes them into the class of irrational behaviour that gives rise to market imperfections that either cancel out or that can be exploited by the savvy investor. Even advocates of behavioural finance sustain that emotions are mostly an annoying impediment to decision making whose likely effect is to thwart the already limited cognitive abilities of investors and traders.

Market practitioners and news commentators routinely employ concepts borrowed from psychology to frame and also to make sense of market events and collective behaviour, and emotions, mood, and sentiments often take on a central role here. A burgeoning empirical literature that provides evidence on investor behaviour and market anomalies is witness to the importance of moods and emotions such as optimism, anxiety, hope, and regret in routine financial decision making. More so, emerging disciplines such as the social studies of finance, naturalistic decision making, and neurofinance have started to make inroads on the difficult topic of emotions, adding crucially to our understanding of how emotions relate to cognition, of their enabling function in affective communication, and of their role in shaping institutions and cultural practice.

Emotions as an enabler of 'good' decision making - individual aspects
Perhaps the most important lesson to take away from these new developments is that the widespread view of emotions as an inhibiting influence on cognition is both incorrect and misleading, and that affective engagement with self and the environment is key to successful decision making. Basic emotions such as fear are adaptive responses conditioned by evolutionary pressures and are critically important for generating quick responses in perilous situations, when a premeditated assessment of options would entail existential risks and thus threaten survival. What is more, recent evidence from research with brain-damaged patients (Damasio, 1994) has demonstrated that emotions are an essential element of effectual decision making. Without them, individuals would get caught up in details, be unable to focus on the task at hand or prioritise the available options, and, in fact, be incapable of deciding at all.

The social aspects of emotion
Above and beyond these limitations which affect the individual in her capacity to make ‘good’ decisions, emotions are key to effective social communication because they help resolve the important problem of double contingency (Luhmann, 1984). That is, they allow socialising individuals to resolve the deep uncertainties about the intentions of others through fast, affective communication, use emotional cues to make their action tendencies known to others and, finally, to gather and interpret emotional cues from the environment. In extremis, emotions shape expectations in social exchange, institutions, and culture – including the culture of rational decision making in firms – and are in turn shaped by these (Pixley 2002). It is here that some criticism is raised by social scientists and cultural theorists (Petta and Trappl, 2001) at the perceived bias of emotion research towards intra-personal characteristics and aspects of affect, when emotions at higher levels of social reality are constructed via rituals, customs, and institutional practice. Needless to say that these aspects, though thoroughly analysed in a small number of well-known ethnographic market studies (Abolafia, 1996), have had so far no repercussion in current financial modeling.

Emotions in artificial intelligence
The emergence of affective computing (Picard, 1995), with its focus on believable emotion generation and recognition in human-computer interaction, has placed artificial intelligence firmly on the map of emotion research-active disciplines. For an important part of the AI community, nowadays it looks "unlikely that appreciable intelligent autonomous behaviour could be achieved without emotions" (Petta and Trappl, 2001). But what just is the relevance of AI for finance in the context of emotion modelling? For one, many market and investment models from financial economics and behavioural finance start out from a (usually homogeneous) population of consumers or investors who maximise their expected utility over the range of available choices. In that sense, they are homogeneous agent-based models which belong to a much broader class of models derived from multiagent systems, the substrate of distributed artificial intelligence (DAI). An important strand in DAI research focuses on emotions and tries to understand how evidence from brain research can be usefully deployed in artificial agent populations for instance to increase believability of virtual characters or to improve coordination in groups of agents. Thus, more sophisticated models of emotion and cognition could be made available through DAI for financial research, but with the need (or possibility) of formulating the market or decision model in the framework of a multiagent system rather than within the traditional deductive paradigm. Developments in emotion modelling within DAI have the further potential to cover a wide range of applications in financial decision training and support, as well as in marketing.

Theories of emotion
Mathematical and computational models of emotion typically range from simple, valence-based models that generate certain actions based on the perceived strength – or valence – of one or several ‘basic’ emotions such as fear or sadness to complex models based on multi-level theories that can process emotion on different levels and modulate for instance processes of perception, attention, and appraisal. While most attempts to incorporate emotions in standard finance theory draw on very basic theoretical accounts or just on common sense, computer models used in artificial intelligence applications are usually grounded in well-established cognitive theories of emotion and draw on a much deeper understanding of how emotion is intertwined with cognitive processes. Still, where less psychological fidelity is needed such as in video games, agent behaviour based on simple hedonic scripts that map situational parameters into an affect response are usually sufficient to generate believable behaviour of artificial characters.

There seems currently to be a preference within the AI community to adopt a cognitive perspective on emotion, which views emotions in broad lines as a result of a more or less complex cognitive evaluation of an internal stimulus or external event. Among the most favoured approaches perhaps is the family of cognitive appraisal theories (Frijda, 1987; Lazarus, 1991), which view emotions as typically adaptive and as directly serving individual autonomy, supporting social communication, and enabling individuals to cope under resource scarcity. Appraisal theory links event stimuli to (re)action through a causal chain of appraisal, affect response, and action tendency. Individuals are continually faced by the prospect that an external event may alter an important feature of her relationship with the social or physical environment. Once such an event has taken place – or possibly in anticipation to it – the individual engages in a process of appraisal to weigh the significance that the event might have for her current interests or concerns, which might give rise to an affect response at different levels of emotion processing. In the second stage, the individual then has to decide on a coping strategy that allows her to manage that new situation, which in turn results in action readiness.

Implementing emotion models
Despite widespread reports to the contrary, financial economists have made well-publicised efforts to incorporate aspects of emotions and moods into standard economic models. Most of these attempts however have been decidedly less ambitious than in AI, either completely ignoring findings from emotion research or using more ‘shallow’ approaches that are further watered down when embedded in the very general utility formulation from economic theory. Emotions in financial economics are evaluated in terms of costs and benefits as is any other decision variable, and there is no room in these models for emotions to take on the more complex roles propounded by contemporary emotion research.

Causal models of emotion such as those espoused in cognitive appraisal theory give us some hope that more advanced approaches to account for emotions and emotion-regulation might enable us to address the limitations of the purely cognitive-theoretical approximations from financial economics. This might be possible within the analytical framework of standard theory or some extension thereof, or alternatively has to be translated to a multiagent framework where analysis is carried out via simulation. As an example, we might consider a model from the family of cognitive appraisal theories proposed by Ortony, Clore, and Collins (1988). The appraisal process focuses on a given range of goals, standards (e.g. ethics), and preferences and results in a combination of 22 valenced emotions that can be incorporated in a utility formulation. The utility function would combine the 22 emotion factors into a personality-dependent decision variable that can then be used in a decision model such as maximum utility or least regret.

To implement a more complex emotion model one might opt for the flexibility and theoretical neutrality of a multi-agent system, where implementation might be more straightforward and shorten the time-to-analysis. In contrast, the mathematical-analytical framework of financial economics imposes a number of important constraints on the modeller such as the requirement for analytical tractability or for utility maximisation. In a multiagent model, individual agents process internal emotional states in line with the injunctions derived from the underlying social or cognitive psychological theory chosen by the designer. The emotion model might be complemented by a personality model that incorporates more stable individual traits such as openness or neuroticism which can modulate emotion interpretation and response.

Individual agents are situated in an physical and social environment in which they interact with other agents and which is the source of external stimuli that flow into the appraisal process. The emotional response generated by the appraisal results in action tendencies that are communicated to the environment through affective channels if they are present. Finally, a decision models selects coping strategies and actions based on the emotion response and character traits. These range from simple scripted models in which agents select actions based on emotional states and personality to sophisticated distributed planning algorithms that choose the appropriate course of action (Arnold, 2003). Some advanced models assume a body focus that permits agents to physically express their emotions in line with the coping strategies they have selected. In even more complex inter-agent models, agents are equipped with the competence to infer emotional models of other agents and to subsequently extract valuable information from the other agents’ observable affective state.

Why modelling emotions
Modelling emotions in financial decision making is important for a number of reasons. First, there is ample empirical evidence from financial markets that emotions have an important effect on decision making and that the traditional view of emotions merely obscuring sound judgment is mistaken. Second, sophisticated simulation and game-based training in financial decision making will require a believable engagement of the learner with virtual characters, which is unlikely to succeed without good emotion recognition and emulation. Moreover, effective decision support might require emotional profiling for which a good model is a prerequisite. Third, large-scale financial infrastructure of the future is likely to be managed by agent systems, which could be made more robust incorporating coping and socialisation strategies explored in emotion research.

Emotions in financial markets
There now exists extensive evidence from neuroscientific research that emotions not only have an important effect on decision outcomes, which might be considered a truism, but that emotions are closely intertwined with cognitive processes to the extent that they are seen as a key facilitator of effective decision making (Damasio, 1994). One implication of this is that the shallow models of negative emotion found predominantly in behavioural approaches to portfolio theory and other areas of financial decision making ought to be revisited and possibly modified in light of this evidence. It is quite possible that many phenomena where decisions are strongly affected by moods, sympathies, panics, and so on cannot be explained satisfactorily with a purely rational-cognitive model of decision making or social exchange. Rather, one will need to carefully gauge whether an explanation coached in terms of rational expectations or judgemental biases truly captures the nature of the problem.

That emotions play a role in financial markets and decision making becomes patently clear if one surveys the empirical evidence found in the financial literature on how sentiment, mood, and personality influence decision behaviour. For instance there is ample support for the assumption that individuals are more focused and perform better at certain cognitive tasks if they are in a positive mood. Hirshleifer and Shumway (2003) demonstrated that on sunny days – which presumably induced an upbeat mood in investors and traders – had a positive effect on stock performance in a number of important US stock exchanges. In another study, Kamstra, Kramer, and Levi (2003) present international evidence that seasonal depression, which correlates with the length of the day, has a negative effect on stock returns. Another strand of this literature contends that the influence of market sentiment, based on the aggregate emotional state of investors in a particular market, exerts a critical influence on the evolution of market prices. A recent empirical study involving a large number of traders at several important investment banks sheds light on the role that emotions play in decision making, stressing the positive role of emotions for highly experienced traders and the importance of emotion regulation for trader performance (Fenton-O’Creevy, 2008 – quotable?).

Decision training and support
Quantitative modelling is an important aspect of financial theorising since it allows a much more rigorous exploration of the implications of a set of hypotheses than a verbal account could do. However, models have many other uses that are not explanatory but rather descriptive in character. For instance they are a key component of simulation and game software used for the purpose of training or decision support, an area that has recently received a great deal of attention from the learning technology and serious games literature (Gunter, Kenny, and Vick, 2006). Decision training software that deploys virtual humans to engage learners in interactive scenarios have to consider how the display of emotions affect the believability of the virtual characters for instance in a negotiation scenario. Interfaces to such training applications will have to be able to process the emotional states of the learner to generate personality profiles and to provide the necessary affective cues to the virtual characters. Emotion models can also be put to use in decision support systems for investors, for instance in a trading platform to assist in security selection and risk management. Here, a OCC-type computational model of emotion might elicit investor goals, preferences, standards, and personality traits and simulate the appraisal process to suggest coping-based trading strategies that are compatible with the investor’s emotional profile.

Multiagent-based simulation
The mathematical-analytical framework of financial economics puts some (legitimate) constraints on what can be achieved in terms of model scale and modelling scope. The flexibility and theory neutrality (at least with respect to economic theory) of multiagent systems forestalls the compromises known from financial economics and allows the designer to implement a model that is true to the specification implied by a particular theory of emotion. Now, what mathematical deduction is for financial economics, simulation is for a multiagent model. Thus the aim of multiagent-based simulation (MABS) is to ‘execute’ a given model formulated in the framework of a multiagent system, by ‘running’ the model over a range of parameter values.

During the 1990’s, MABS has developed into a research field in its own right and presents us with a powerful framework for modelling and analysing micro- and macro-social phenomena such as racial segregation (Schelling, 1978)) or emergent system dynamics (Gilbert, 1995). Since DAI has developed a number of advanced models of emotion and embedded them in a multiagent system, there might be a potential in transferring some of these models into MABS models of financial markets.

Emotions and financial theorising
The traditional rational choice approach to solving the problem of scarce resource allocation developed in economic theory represents one of the cornerstones of modern financial economics. Despite the fact that a number of important anomalies observed in experimental settings as well as in actual financial markets are hard to explain within an efficient market and rational expectations framework, economists have mostly resisted calls to incorporate insights from psychology into their theorising – it would have the potential to dilute the axiomatic basis and parsimonious appeal of their discipline. As a result, dissidents have embarked on an alternative to financial economics – behavioural finance – which seeks to incorporate evidence from psychological experiments and empirical studies into financial models. Much of behavioural finance is experimental and empirical rather than theoretical, and considerable attention has been directed at laying bare the limitations of human cognitive capabilities and the attendant biases in judgement. At the same time, the almost exclusive focus on cognition as the primary mechanism in decision making comes at the cost of neglecting evidence from emotion research on how affect processes and structures are (1) thoroughly intertwined with deliberate processing and (2) a vital element of competent and effective decision making and socialising. Moreover, behavioural finance shares with the orthodoxy a commitment to methodological individualism at the cost of neglecting the influence of social and cultural structures and processes, and the situatedness of action.

Incorporating emotions in economic theory
Historically, and certainly before the emergence and widespread acceptance of modern axiomatic utility theory, discourse of emotion played a more central role in economic theorising (Bentham, 1948; Keynes, 1937) than they do today. However, and despite the exclusive focus of financial economics on perfect rationality, there have been some noted attempts to explicitly incorporate emotional aspects into economic theory. Much of these efforts however are restricted to particular situations and specific areas of applications, although there are a handful of attempts at a more general integration of affect into financial decision theory (Loewenstein, 2000). In fact, for many economists standard expected utility theory (EUT) is sufficiently flexible to cover any scenario of choice, and emotions such as anxiety or guilt can, according to this view, be incorporated directly into the utility formulation the same way as you would do with any other consumption item (Hermalin and Isen, 2000). To be fair on this point, critics from within financial economics concede that it might be difficult if not impossible in practice to form coherent preferences over emotions in particular since different emotions might not even be comparable among each other (Merkle, 2007).

Interplay of cognition and emotion
Economists might be right to claim that standard EUT is well equipped to incorporate non-material aspects such as fear, regret, happiness and so on. Unfortunately, such a model would not be able to account for the multifaceted influences that emotion can have on perception, preferences, appraisal, memory retrieval, and so on. The standard utility model sees emotions in the same terms as any other element to which utility can be assigned to, and, in the simplest case, as merely contributing to utility via their hedonic value or valence. In reality however, emotions, both immediate and anticipated, exert more far-reaching influences, influencing perception of other costs and rewards, impacting the way we search and select meaningful options, affecting recall of past events, colouring context, and altering the expectations of future events. Thus, even in a simple model of decision making, emotions cannot easily be thought of on the basis of 'first principle', but they really should be seen as affecting the decision process at all its stages. Moreover, emotions play a central role in social relationships, where they aid individuals in the process of socialisation and sense making, and where they shape and in turn are shaped by institutional and cultural belief systems. More to the point however, the relationship between emotion and cognition which has recently found much attention in brain research and also in artificial intelligence, is of critical importance if we are to satisfactorily explain emotionally ‘charged’ phenomena in finance, something that could be achieved by formulating financial models for instance in terms of cognitive appraisal theory to account for the causation of emotions through cognitive processes and also the way how emotions feed back into the cognitive appraisal process.

Behavioural finance
Behavioural finance has mainly focused on the cognitive limitations of human decision makers, as a response to the unrealistic assumptions of omniscience, perfect rationality, and unlimited cognitive abilities that underlie models in neoclassical finance. Moreover, drawing on psychological research, behavioural finance intends to explain a wide range of well-known anomalies in individual behaviour and in financial market prices. Much of the work in this area is experimental and empirical, with emotions entering the picture only by name rather than by existing theories or empirical evidence from emotion research. Behavioural finance explains the prevalence of anomalous, persistent price patterns by a simultaneous presence of judgemental biases of investors – which might be emotion related – and the limits to arbitrage (Shleifer and Vishny, 1997) that makes it impossible in some cases for more sophisticated market players to exploit deviances of prices from levels predicted by the efficient market hypothesis.

Proposed by Kahneman and Tversky in 1979, prospect theory is one of the earliest and perhaps also the most influential behavioural finance theory to date. Emotional attachment of individuals to their current level of wealth is expressed as regret, an anticipated feeling of disappointment, that leads to decision strategies aimed at avoiding the negative feeling that goes with the prospect future losses. As a consequence, investors become risk loving in the face of losses and are for instance reluctant to close out a loss-making position. In fact, the dread of realising a paper loss engenders a preference for high-risk strategies based on wishful thinking for a hypothetical future recovery from current paper losses, as has occurred in some well-publicised cases of rogue trading (Leeson and Whitley, 1997). A similar criticism to neoclassical theory can be raised with respect to the way that emotions are dealt with in behavioural finance. In fact, by focussing essentially on the cognitive psychological shortcomings of EUT, proposals for theoretical models in behavioural finance leave important aspects of emotions, distributed cognition, and culture largely unaccounted for.

Having said this, behavioural finance has produced a large body of empirical evidence that corroborates the view that emotions are an important ingredient of financial decision making and, in some noted cases, have a lasting effect on price patterns in financial markets. Investor sentiment such as optimism or depressive mood has been shown in some instances to have a noticeable effect on stock market performance (Tetlock, 2007; Hirshleifer and Shumway, 2003; Kamstra, Kramer, and Levi, 2003). Overreaction (or underreaction) to new information, for instance on quarterly company results, has been explained with reference to a range of judgemental biases and decision heuristics (Barberis, Shleifer, and Vishny, 1998; Daniel, Hirshleifer, and Subrahmanyam, 1999). Interestingly, it is not only less sophisticated investors whose decision process is marred by cognitive shortcomings. There are notable instances of misjudgement commonly found in professional traders and investment managers. Besides, emotions such as dread, fear, or anxiety are clearly relevant in terms of risk perception and can induce quite dramatic changes in the level of risk that private investors, traders, and fund managers are prepared to take, thereby not only affecting probability judgements, but also potentially leading to a reframing of the extent and impact of potential dangers and other future contingencies. However, despite the important empirical evidence vouching for the relevance of emotions in finance both at an individual level and in social interaction, behavioural finance pays mostly a lip service to the new and more complex theories emerging from emotion research, a weakness that is likely to be tackled by the merging discipline of neurofinance.

Examples of emotion models
In some noteworthy occasions, finance scholars have put forward models of emotion formulated in the framework of or as an extension to standard expected utility theory. While many models proposed in behavioural finance can be interpreted in terms of emotion expectations or anticipations such as anxiety or anticipation of regret, there have been some efforts to model the effects that actual feelings and moods have on the decision process.

Visceral factors
Loewenstein's model of immediate emotions, or visceral factors, is a very general model based on an inter-temporal utility formulation, where visceral factors are represented as state-dependent preferences (Loewenstein et al., 2001). The author contends that standard models of utility that do not appropriately account for emotional factors, have great difficulties to explain observed behaviour where such factors play clearly an important role. Interestingly, the author cautions against the common assumption that behaviour is necessarily the product of deliberation, which, given the standard assumptions of perfect rationality, is quite surprising and perhaps a little off the mark. Loewenstein points to three areas in economic behaviour where visceral factors are thought to play a particularly important role: bargaining and negotiation, inconsistencies in inter-temporal choice witnessed by the extreme discounting of future expected pain, and the effect of emotions in the evaluation and perception of risks. An encouraging aspect of the article is that it embraces the idea that emotions can have a positive influence on individual well-being and survival, though always with reference to utility maximisation.

SP/A theory and BPT
Experimental research by Lopez (1987) has led to a more specific proposal of how the basic emotions of fear and hope might help to explain for instance why individuals who buy insurance also buy lottery tickets – known as the Friedman-Savage paradox. Lopez’s theory builds on the so-called safety-first principle, which echoes a general concern for survival by setting a minimum subsistence level for consumption in choice situations. Lopez proposes a new approach, SP/A theory, that extends the concern for security and its correlate of 'fear' to include aspiration, which is related to 'hope'. Fear and hope simultaneously serve to alter the probability weights in the calculations of expected wealth, with fear reducing and hope increasing expected wealth respectively. This framework has been used as a basis for behavioural models in portfolio theory (Shefrin and Statman, 2000), where investors take decisions with a view to secure a minimum subsistence level while at the same time invest to satisfy higher aspiration levels of wealth.

Other models
There have been many more attempts at addressing emotions directly within standard utility theory or drawing on more recent theories from behavioural finance. Caplin and Leahy (2001) have proposed an extension to EUT they call psychological utility theory where anticipatory emotions such as anxiety or suspense are represented as additional state space variables that have an impact on time-dependent preferences. As with many other alternative utility formulations, the goal is to explain anomalous but prevalent market phenomena such as the equity premium puzzle that are not well captured by standard approaches. The authors propose a model for portfolio selection that incorporates time-varying anxiety, an apprehensive feeling in the face of an uncertain future - not to be confused with the static concept of risk aversion - to explain certain investor behaviour in terms of anxiety avoidance. In a similar vein, Mellers, Schwartz and Ritov (1997) have built on decision affect theory, a blend of regret and disappointment theory, to propose a new offspring of utility theory – subjective expected pleasure – to guide financial decision making such as portfolio choice in the face of hedonic influences. A model that is motivated by cognitive appraisal, a popular family of theories that link emotions to coping tendencies, is Lerner and Keltner's appraisal tendency approach. Finally, the affect heuristic proposed by Slovic et al. (2002) has been used to explain why the affective image of a company or a sector gives rise to abnormal returns of their shares (Ganzach, 2001).

Social mood
Both neoclassical and behavioural finance have made important strides in explaining investor behaviour on an aggregate level. The stakes here are high, since speculative bubbles, fads, and market panics which are seen as a phenomenon of mass psychology have important implications for systemic stability and thus for policy and oversight. It is generally assumed that emotions lie at the heart of the mechanisms that temporally cause uniform behaviours in the market place which in turn may lead to abrupt price changes. Prices for instance are likely to act as an external stimuli that affect or induce what is known as investor sentiment (Delong, Shleifer, Summers, and Waldmann, 1990). This in turn impacts on market performance through an expectation dynamics generated by hedonic evaluation and emotion anticipation. Herding is another phenomenon that is thought to be triggered and fed by emotionally coloured communication or social observation. More generally, empirical studies have shown that other factors might affect social mood (Nofsinger, 2005) with important consequences for aggregate investor behaviour and market performance.

Recent developments in emotion research for financial decision making
The fact that the bulk of quantitative emotion models for financial behaviour and decision making have been proposed by neoclassical and behavioural finance scholars shouldn't conceal the many insightful contributions to the understanding of the role of emotions in financial decision making by other disciplines such as expertise research, social studies of finance, and neurofinance. In contrast to the mathematical orientation of financial economics and the predominately laboratory-based experiments of behavioural finance, both research in expert decision making and the sociology of financial markets have a strong empirical orientation in the tradition of the observational methods espoused in anthropology and cultural studies, and stress the importance of attending to situation and context when explaining individual or market behaviour.

The role of emotion and its impact on trader performance has been extensively analysed and reported by Fenton-O'Creevy and colleagues (2008?) in a large-scale study that involved 118 traders from four leading London-based investment banks. They found that successful expert traders are more likely to engage in self-monitoring and in emotion regulation, and that this carries important benefits in terms of trading performance. The implications in terms of current efforts in financial economics to model emotions are that (1) simple valence-based models cannot capture the regulatory dimension of emotions and (2) traders are not a uniform bunch of people, but show significant differences in the way they process information and deal with emotions, which is likely to have a significant impact on trading style, strategies, and performance.

In an influential ethnographic study of stock, bond, and futures markets, Mitchel Abolafia (1996) shows that the image created by finance theory of an efficient market engendered by hoards of independent individuals striving for their own profits is quite mistaken. Traders and investors are in fact caught up in a network of social expectations, cultural scripts and symbols, and political power relations, constantly negotiating between the extremes of pure self-interest and institutional self-restraint. Emotional engagement fostered by market-specific rituals plays an important role in instilling strongly-held views of self and community, and thus provides a fertile ground for self-restraint. A model framework that is purely based on principles of methodological individualism will find it difficult if not impossible to model social and cultural structures as well as processes that mediate between these structures, group-level emotions, and individual cognition.

Despite the many insights delivered by behavioural research into financial decision making, theories that explicitly address emotions stand, like those of their neoclassical counterpart, often on very weak theoretical and empirical grounds, having to rely on evidence from observable behaviour and self-reporting, which is likely to provide, in particular for the complex phenomenon of emotion, no more than a patchy image. Neurofinance, a new research discipline that seeks to understand financial decision making in terms of the underlying neural processes, provides us with a new tool kit that goes some way in addressing the shortcomings of behavioural experiments and to shed light on whether theoretical constructs such as risk aversion, regret, or utility have a neural basis. The persistence of stock market bubbles for instance can be explained by a self-reinforcing process which generates positive somatic states. Or one might ask the question of how the brain encodes information on risk/reward structures. In a fMRI experiment, Preuschoff, Quartz and Bossaerts (2006) analyse whether a utility or a mean/variance representation is used by the brain in the context of portfolio selection. Emotion models, which are at present rather ad hoc in financial theory, might be adapted to reflect new evidence coming from brain studies, in particular as to the interrelation between emotion and cognition.