Appraisal

A Cognition Briefing

Contributed by: Carlos Herrera, University of Ulster

Introduction
In this briefing we consider appraisal as a cognitive phenomenon of importance for adaptation in autonomous agents, as well as essential for the development of higher-order cognitive capacities in humans. Appraisal is a concept used in many different areas, such as economics, decision making, real state or education, and which is defined in the Encarta dictionary as “a judgment or opinion of something or somebody, especially one that assesses effectiveness or usefulness”.

In cognitive science, Magna Arnold (1960) introduced the notion of appraisal as the process through which the significance of a situation for an individual is determined. Since then, this notion has been central in theories of emotion (e.g. Frijda 1986, Lazarus 1991, Scherer et al 2001). Emotion is the response to significant or concern-relevant situations - in this context appraisal refers to the process that allows agents to asses environmental pressures and elicit responses. Appraisal theoriestake into account that emotions show awareness of relevance of situations.

Appraisal is essential for autonomy insofar judgements of value underlie adaptive responses. The capacity of responding to situations on the basis of how they concern the agent’s adaptation is essential for coping with unavoidable changes in the relationship with the environment. The discriminatory power of appraisals (in direct relation to the complexity and richness of the ecological niche and the range of situations adapted to) can be a world apart in simple organisms and humans, but its central role in adaptation shares some fundamental characteristics.

Despite the importance of appraisal for adaptation, the use of the concept is mostly constrained to emotion theory and lacks development in wider areas of cognitive science, especially in robotics, where other categorisation tasks take prevalence. Nevertheless, appraisal can be seen as a fundamental cognitive phenomenon, arguably more basic than other categorization tasks. This briefing aims to provide a general introduction to the notion, stressing the importance for the development of autonomous robots.

Appraisal and Higher-order Cognition
Critiques of appraisal theories are often built on the assumption that appraisals can only follow higher-order cognitive processing. For an overly cognitivist standpoint, appraisals would be the end-point of complex cognitive processing: "emotional event elicit a number of cognitive appraisal steps, in a fixed or variable sequence, or perhaps in parallel ... after the appraisal steps, emotion follows as their result" (Frijda 1993). Facts and issues of the situation are evaluated independently to produce an overall evaluation. Emotion is special because it takes control precedence over other systems when rough evaluations have been made. In a way, we need to know a lot about things before we can know what they mean to us.

This view is nevertheless far from representing appraisal theorists. Maybe at the root of the misunderstanding is that the notion of appraisal originated in the study of human psychology. In humans, emotional processes may be simple, triggered by association and clinically treated with through psychiatric drugs. Nevertheless, they reflect complex appraisal patterns, as our everyday emotions mostly concern how we interpret our social context, which is cognitively intricate. What is relevant for human agents, apart from specific biological concerns, lies in the social context and is mediated by complex articulations and interpretations of self and others. Other cognitive agents, such as animals, lack such complex framework, and appraisals are more immediate and concerns more direct.

Even in the case of appraisals that relate to complex social structures, actual appraisals often occur in an unconscious and automatic fashion (Lazarus 1991). As by outlined Frijda (1993), much of appraisal theory and research suffers from confusing these two roles of appraisal: those of content and of antecedent. In short, the complexity of the content of appraisal need not be matched by a complex appraisal process and vice versa. The simple sight of a swastika can carry very complex negative appraisals, while very complex appraisal process can be dedicated to simple matters such as the choice of sits in a cinema (specially in people exhibiting schizophrenia).

Widening the scope of appraisal to animals and also robots, we may consider appraisals as phenomena that occur in system-environment interaction. Without falling into behaviouristic reduction, a system can be said capable of appraisal if it can produce distinguishable behaviours in response to perceived differences in present or projected value.

In other words, appraisal is as emergent phenomena of interaction, rather than an internal mental processes.Reaction to single stimuli (such as sugar gradient for lower species, or the smell of cats for a rat) may be sufficient for the system to appraise and respond to the value of a situation. The world nevertheless may press for more refined appraisals, as for instance discrimination of danger from benign conditions (as observed in fowls and fish, Tinbergen 1951). In human environments, single stimuli are almost never sufficient to grasp the significance of situations.

Appraisal as a Cognitive Phenomenon
Appraisal, considered a cognitive phenomenon, have some properties that allow distinguishing it from other cognitive phenomena. One characteristic that is often mentioned is the role physiological changes and feeling plays in the appraisal process, which may be considered “hot cognition” (e.g. Zajonc 1980), in contrast to objective knowledge or “cold cognitions”. Appraisal can also be distinguished in terms of cognitive content. For instance, we can identify appraisal as subjective knowledge. While the content of objective knowledge are properties and states of processes and entities independent to the cogniser, the content of appraisal are concern-relevant features of agent-environment interaction . We must be aware that rational processes are never sufficient to bring about appraisal. If we understand fear as the appraisal of danger (well accepted view, e.g. Frijda 2007), rational processes may convinces us all to believe that a first event (e.g. smoking) is more dangerous than another (e.g. flying), while for some people only the second is appraised as fearsome. Or, we may know that we are well prepared for an exam, but still expect the unexpected. In evolutionary terms, appraisals are not selected for providing objective truths, but for how they prepare us for action. Appraisal is a form of perception as it facilitates new knowledge of the environment. The following distinction belongs to Lazarus (1991): primary appraisal allows us to know that the situation is relevant, secondary appraisal asses the ways in which the situation can be coped with. Appraisal can also be strongly linked to association mechanisms – evaluations are easily transferred throughout situational factors. Rolls (2005) argues that instrumental reinforcers (punish or reward) are the processes that, through association to stimuli, elicit appraisals – though, does punishment and reward already imply the positive/negative appraisal of the event? In turn, most forms of associative learning involve some degree of appraisal. In summary appraisal is a cognitive phenomenon that refers to judgements of value. This implies that we refer to a judgement as appraisal if it targets qualitative descriptions of objects or events. Rather than asking “what are the inherent properties of this object or event?”, appraisal asks, “how does this situation (object or event) affect my situation as an agent?”.

Appraisal and Emotion Elicitation
Appraisal has mainly been studied in the context of emotion elicitation, as an antecedent of emotional response. Emotions can be said to be caused by appraisals. Nevertheless, it is important to clarify issues of causality (Lazarus 1991, chapt. 5) – does it mean that the response process can only follow an independent appraisal process? While most studies attempt to isolate appraisal factors, there is enough evidence to suggest that appraisal and response share many neural and physiological mechanisms and cannot be understood independently. A systems approach is required to allow reciprocal causality (Lazarus 1991) – in the emergentist paradigm, as well in dynamical system approaches (Lewis 2005) appraisal and response can be considered as emergent phenomena of the interaction of multiple physiological, neural and situated processes.

States of arousal, for instance, are a strong factor in appraisal (Schachter and Singer 1962) as well as in the generation of response (as states of action readiness, Frijda 1986). Similarly, facial and postural expression can condition appraisal (e.g. Laird, 1984), but are themselves modes of interaction … “they do not express modes of contact. They are modes of contact." (Frijda 1986, p.12). The embodied appraisal process

We have seen how the content of appraisal refersnot to objective properties of external objects, but always to relational features (importance, worth, significance, usefulness). This is closely connected to the question of how appraisal is achieved – rather than through a rational process of inference (detachment and disengagement) – appraisal is embodied and situated. Appraisal is rooted in the actual significance things have for us, what possibilities for interaction and adaptation are open, and ultimately in how make us feel.

Rational processes, whenever they occur, may play an important role in how we feel (and most of the time they are key to understand appraisal), but there is also a strong component of “gut feeling”. This suggests that physiological systems may play an essential role in the appraisal process. Damasio has recently presented investigations that stress the role of physiology (e.g. hormones, Damasio 1994) and physiology related neural structures (e.g. body maps, Damasio 1994), in the management of information to assist cognitive tasks (e.g. anticipation based on as-if-loops). Similarly, the theory of “embodied appraisal” (Prinz 2005), is based on the hypothesis that an essential process in appraisal is the detection of states of the body that, represent “core relational themes” – relations between an organism and its environment that pertain to well-being.

One open research question is by virtue of what physiological states may represent relational features. An interesting factor is the close relationship between appraisal and response. The “goal” of appraisal is not to make use have more detailed information about the world, but to prepare us for situations. Physiological states can be a decisive factor in whether we appraise a situation as dangerous or not, but more importantly it is a decisive factor on how we react to the situation. Arousal, meaning the coordinated physiological responses that underlie an emotion, are states of readiness for action and response (Frijda 2007). The actual significance of internal states in current relationships with the environment and the relational dynamics may be considered a central factor to explain the correlation between the internal and the external.

Appraisal in Robotics
Work in cognitive robotics has been profoundly transformed over the last decades. In classic robotics the aim was to implement a cognitive program (written offline) into a robot. Despite the overwhelming difficulties of real-world interaction, the cognitive abilities of robots where conceived as deriving from AI techniques, dealing with abstract representation through inference methods. In embodied cognitive robotics cognition, rather than predefining behaviour, emerges from the interaction of multiple components, such as neural networks, morphology and environmental situatedness, or more generally sensory-motor coordination. An un-structured stream of sensory information has to be integrated, not only in neural terms but also in behavioural (for instance wall following), in order to facilitate cognitive functions in autonomous behaviour.

This implies that a task within cognitive robotics is to identify relevant embodied cognitive functions, and propose different methods to synthesise such functions.In practice, research is conducted by identifying relevant mechanisms (e.g. neural network architectures and procedures) and investigating ways in which they can be used to control cognitive tasks. Despite the emphasis on grounding, much research still focuses on representational cognition, such as discrimination of objects, map building or planning. There have nevertheless been efforts to consider forms of cognition that depart from classic representational paradigm. In what we may call affordance-based robotics cognition is fundamentally the recognition of affordances for action. In other words, the content of affordances is not an objective property of an entity, but a relational property in terms of what can be done. Work in robotics is increasingly interested in exploiting the notion of affordance [http://www.eucognition.org/wiki/index.php?title=Affordances:_The_review_of_an_inspiring_notion (briefing on affordances) ]. Pfeifer for instance develops a robot that can discriminate between objects based on whether they can be grasped or not (Pfeifer and Scheier 2001): perception of affordances can be seen as functional appraisal. A different approach is the stress on autonomy, epitomised in the fungus eater approach (Toda 1982) In this approach, robots need to be capable of continuous operation over indefinite life-cycles. A basic requirement for this is energy maintenance, which involves behaving differently in different circumstances, being these determined both by internal states (e.g. energy level) and environment (availability of resources). Research on energy autonomy and internal motivations therefore also investigates a form of appraisal. Work on emotion-based robotics is naturally concerned with the capacity for appraisal of concern-relevance. An essential role for emotion-based robotics is to design systems that are capable of categorizing situations that may be significant for the operation of the system. Such appraisal must be expressed in the change of relation with certain aspects of the environment towards behavioural adaptation. It is common, nevertheless, to consider appraisal and response as independent processes, modelling how a cognitive process of appraisal can trigger a behavioural change. With this aim, appraisal is often simplified through the use of quasi-labelled worlds (e.g. dangers are marked with a particular colour). Real progress will be made when appraisals are appropriate in highly ambiguous situations, and express clear cognitive capacities such as anticipation.

References
Arnold M.B. (1960) Emotion and personality (Vols I & II) New York, Columbia University Press.

Damasio, A. (1994). Descartes’ error: Emotion, Reason and the Human Brain. Picador, Cambridge, MA, USA.

Damasio AR: The Feeling of What Happens: Body and Emotion in the Making of Consciousness, Harcourt Brace, New York, 1999

Frijda, N.H. (1993): The Place of Appraisal in Emotion. Cognition & Emotion (7)(3&4) 357-388

Frijda, Nico H. (2007)The Laws of Emotion. Mahwah, NJ: Lawrence Erlbaum Associates, Publishers.

Laird, J. D. (1984). The real role of facial response in the experience of emotion: A reply to Tourangeau and Ellsworth, and others. Journal of Personality and Social Psychology, 47, 909-917

Lazarus, R.S. (1991). Emotion and adaptation. Oxford University Press.

Lewis, M. (2005) Bridging emotion theory and neurobiology through dynamic systems modeling Behavioral and Brain Sciences, Cambridge Univ Press, 2005, 28, 169-194

Parkinson, B. & Manstead, A.S.R. Appraisal as a cause of emotion. In M.S. Clark (Ed) Emotion. Review of personality and social psychology, Vol 13, 1992, pp.122-149

Prinz, J. (2004) Gut Reactions: A Perceptual Theory of Emotion. Oxford University Press, USA.

Rolls, E. (2005) Emotion Explained Oxford University Press, USA.

Schachter, S & Singer, J. (1962) Cognitive, social and physiological determinants of emotional state. Psychological Review, 65, 379-399.

Scherer, K. R., & Shorr, A., & Johnstone, T. (Ed.). (2001). Appraisal processes in emotion: theory, methods, research . Canary, NC: Oxford University Press.

Toda, M. (1982). Man, robot and society. The Hague: Nijhoff.

Zajonc, R. (1980).Feeling and thinking: preferences need no inferences. American Psychologist, 39, pp. 151-175.