Contemporary models of intuition

Psychologists have long distinguished between two modes of thinking, referred to here as ‘deliberative thinking’ and ‘intuitive thinking’ (intuition). The main differences between the two are shown in Table 1. Reference Betsch, Plessner, Betsch and BetschBetsch (2008: p. 4) defines intuition as:

TABLE 1 Characteristics of ‘intuitive’ and ‘deliberative’ thinking processes

This definition does not explicitly include instinctive behaviours, although other theorists accept that intuitive responses can involve ‘a mix of innate and learned behaviour’ (Reference Hogarth, Plessner, Betsch and BetschHogarth 2008).

Central to this definition is the concept of a long-term knowledge store, based on a person’s experience of being exposed to a variety of stimuli and corresponding responses and outcomes. These associations are acquired experientially via observational learning and instrumental and classical conditioning (Reference Epstein, Plessner, Betsch and BetschEpstein 2008). There are two corresponding principles to such learning (Reference Hogarth, Plessner, Betsch and BetschHogarth 2008):

1. • observation of frequencies of events and objects in the environment and the extent to which they co-vary;

2. • the principle of reinforcement: that is to say, the positive (reward) or negative (cost) value of environmental phenomena.

Importantly, the output of the intuitive system is a feeling rather than a verbal proposition. Often this will involve at least a ‘faint whisper’ of emotion (Reference Slovic, Finucane and PetersSlovic 2004), indicating a specific quality such as goodness, badness or riskiness. However, less emotionally charged feelings such as the ‘feeling of knowing’ (Reference Liu, Yanjie and XuLiu 2007) can also occur. Such feelings are believed to help people to navigate quickly and efficiently through the complex and sometimes risky environments of everyday life. They have attracted a variety of terms, such as ‘vibes’ – defined as ‘vague feelings such as disquietude and agitation’ (Reference Epstein, Plessner, Betsch and BetschEpstein 2008).

An influential neurobiological model of decision-making – the ‘somatic marker hypothesis’ (Reference DamasioDamasio 1994) – similarly proposes that feelings arising from a longer-term knowledge store help to guide behaviour. This model posits a long-term knowledge store that links new stimuli to ‘somatic markers’ (internal mental representations marked by positive or negative feelings, linked to distinctive ‘somatic’ visceral states) that guide decisions advantageously and mostly unconsciously.

Research in cognitive psychology has confirmed that, depending on the predispositions of the person and the characteristics of the task at hand, some decisions are made following a predominantly effortful thinking process and others tend to be reached with more reliance on rapid, emotionally coloured ‘gut feelings’ (Reference Deutsch, Strack, Plessner, Betsch and BetschDeutsch 2008).

Most decision-making theorists now accept that optimal decision-making requires an integration of both deliberative and intuitive modes (Reference Baumeister, DeWall, Zhang, Vohs, Baumeister and LoewensteinBaumeister 2007; Reference Plessner, Czenna, Plessner, Betsch and BetschPlessner 2008). The two systems appear to operate in parallel and depend on each other for guidance (Reference Slovic, Finucane and PetersSlovic 2004), with the relative contribution of each system varying according to the situational demands and the person making the decision (Reference Epstein, Plessner, Betsch and BetschEpstein 2008). In practice, we can have reflective thoughts about our feelings and intuitive feelings about our thoughts, such that the two systems interact simultaneously and sequentially (Reference Epstein, Plessner, Betsch and BetschEpstein 2008). Various terms have been coined for the proper integration of both modes of thought to produce truly ‘rational’ decision-making, including the concept of ‘affective rationality’ (Reference Peters, Vastfjall and GarlingPeters 2006) and ‘the dance of affect and reason’ (Reference Finucane, Peters, Slovic, Schneider and ShanteauFinucane 2003).

Problems with intuition in risk assessment

Progress in violence risk assessment in clinical practice reflects the realisation that both modes of thinking have potential value. Methodologies were previously limited to either purely unstructured approaches (which gave clinical intuition free rein; Reference MonahanMonahan 1984) or to hyperrational actuarially based models (which sought to minimise the role of intuitive clinical judgements; Reference Quinsey, Harris and RiceQuinsey 1998). However, there is a third way, generally called ‘structured professional judgement’, that attempts to integrate the best features of each approach and has been shown to be both reliable and valid (Reference Douglas, Reeves, Otto and DouglasDouglas 2010b).

The emerging consensus in the decision-making literature is that integrated models incorporating both deliberative and intuitive reasoning are most appropriate to meet real-world challenges. However, the precise role of the intuitive system remains underdeveloped in the context of violence risk assessment. There are several possible reasons for this (Box 1).

Guidelines for using intuition

The intuitive system can lead to dangerously biased and prejudicial decisional outcomes. It is therefore unsurprising that pure unstructured clinical intuition is now rarely advocated as best practice for violence risk assessment in mental health. However, the widely advocated risk assessment approach of ‘structured professional judgement’ (Reference Webster and HuckerWebster 2007) must perforce incorporate the professional’s intuitive functioning. Therefore, the question is not ‘Can intuition play a role?’, but ‘How can the role of intuition be best utilised?’.

Ideally, deliberative and intuitive modes need to be carefully synthesised with relative emphases and inputs appropriate to the decision or task at hand. Preclinical laboratory-based research suggests that ‘either system can out-perform the other depending on the nature of the problem at issue’ (Reference Epstein, Plessner, Betsch and BetschEpstein 2008). Therefore, the task is how to apply reason to temper strong emotions engendered by some risk issues and how to infuse ‘needed doses of feeling’ into circumstances where lack of experience may result in decision-making that is too ‘coldly rational’ (Reference Slovic, Finucane and PetersSlovic 2004).

The structured professional judgement literature (e.g. Reference Webster, Douglas and EavesWebster 1997; Reference Douglas, Webster and EavesDouglas 2001; Reference Webster, Martin and BrinkWebster 2004) and associated training workshops (e.g. Reference OgloffOgloff 2011) tend to focus largely on the ‘structured’ aspect – the operationalising and rating of empirically derived risk factors – but provide less guidance for the ‘professional judgement’ aspect. The following section attempts to provide such guidance, based on proposals developed by Reference HogarthHogarth (2001), a cognitive psychologist who has extensively researched the functions and limitations of intuitive thinking. Essentially, his proposals comprise a set of guidelines by which the deliberative system may systematically be utilised to guide the intuitive system, helping to optimise overall decision-making processes. Each guideline (Box 2) will be considered in the context of violence risk assessment.

Guideline 2: Use your own emotions as a source of data

Feelings or ‘vibes’ generated by the intuitive system, Reference Hogarth, Plessner, Betsch and BetschHogarth (2008) asserts, can usefully be treated simply as part of the informational matrix to be considered in any decisional context: ‘rather than ignoring or trusting one’s emotions blindly, I believe it is best to treat emotions as data [that are] just part of the data that should be considered’.

In the context of violence risk assessment, it is useful to consider two distinct kinds of such data to which the clinician will have access: emotions about patients that arise in the context of the therapeutic relationship; and emotions about the accuracy or otherwise of the clinician’s own risk assessment. These will be considered in turn.

Interpersonal dynamics in therapeutic relationships

A basic tenet of psychodynamic therapeutic practice is that clinicians can develop a deeper understanding of their patients by using their own emotions about the patient as a guide. Such emotions are often considered under the rubric of ‘countertransference’ and awareness of the transference/countertransference relationship ‘allows reflection and thoughtful response rather than unthinking reaction from the doctor’, as this journal has reported (Reference Hughes and KerrHughes 2000).

The academic psychologist Paul Meehl, who is generally best remembered for his strong advocacy of actuarial methods of decision-making, nonetheless recognised that determining the significance of the interpersonal dynamics evident in a clinical relationship is a task that is inevitably principally handled by the intuitive rather than the deliberative system (Reference MeehlMeehl 1954):

‘In a therapeutic handling of the case, it is impossible for the clinician to get up in the middle of the interview, saying to the patient, “leave yourself in suspended animation for 48 hours. Before I respond to your last remark it is necessary for me to do some work on my calculating machine” […] at the moment of action in the clinical interview the appropriateness of the behaviour will depend in part upon things which are learnable only by a multiplicity of concrete experiences and not by formal didactic exposition’ [my italics] (pp. 81–82).

Thus, he recognised that informed clinical experience is essential to a sophisticated understanding and management of interpersonal dynamics in the therapeutic context.

Interpersonal style

Empirical consideration of the role of clinicians’ emotions as data, however, requires a theoretical framework to provide a basis for the operational definition of such emotions. The concept of ‘interpersonal style’, based on Kiesler’s theory of interpersonal behaviour (Reference Kiesler, Schmidt, Wagner, Plutchik and ConteKiesler 1997), may be used to provide such a framework. Interpersonal style refers to how individuals characteristically relate to each other and how people perceive themselves in relation to others. Although it is considered to be a characteristic of an individual, people react to their interpersonal environment and in turn influence the relationships that others have with them. Such interactions can be considered and understood in relation to the core dimensions of ‘power’ or ‘status’ (which ranges from dominance to submission) and ‘affiliation’ (which ranges from hostility to friendliness) (Fig. 1) (Reference LearyLeary 1957; Reference Kiesler, Schmidt, Wagner, Plutchik and ConteKiesler 1997). Behaviour generally elicits a corresponding response from the affiliation dimension (i.e. friendliness elicits friendliness, hostility elicits hostility), but reciprocal responses from the control dimension (i.e. dominance elicits submission and vice versa).

FIG 1 The interpersonal circumplex (after Reference Kiesler, Schmidt, Wagner, Plutchik and ConteKiesler 1997).

An understanding of interpersonal style and of the corresponding countertransference feelings can help clinicians to avoid overly reactive decision-making, which is sometimes prompted by emotional responses to their patients. Such reactivity is generally recognised to lead to an inefficient decision-making strategy in interpersonal settings (Reference Baumeister, DeWall, Zhang, Vohs, Baumeister and LoewensteinBaumeister 2007).

The value of emotional responses

A careful consideration by clinicians of the informational value of their own emotional responses can improve clinical judgement and decision-making. In the context of violence risk assessments, understanding interpersonal style (by tapping into the clinician’s own emotional reactions in the interpersonal encounter) may help to indicate the likelihood of future interpersonal conflict. It may also give clues as to particular dynamic ‘risk signatures’ (as has been explored in this journal; Reference Reiss and KirtchukReiss 2009) or the ‘relevance’ (Reference Douglas, Reeves, Otto and DouglasDouglas 2010b) of specific interpersonal interactions for a particular person. For example, the interpersonal style of a narcissist, marked by ‘dominance’, may be dangerously sensitive to even minor slights, which are perceived as intensely humiliating (Reference NestorNestor 2002).

Research tools

The Chart of Interpersonal Reactions in Closed Living Environments (CIRCLE; Reference BlackburnBlackburn 1998) and the Impact Message Inventory – Circumplex (IMI–C; Reference Kiesler and SchmidtKiesler 2006) are empirical measures, comprising structured questionnaires filled in by clinicians, that can rate a patient’s interpersonal style based on the clinician’s emotional responses to the patient. Assessment of interpersonal style using these tools shows good interrater reliability (Reference BlackburnBlackburn 1998; Reference Kiesler and SchmidtKiesler 2006) and research has suggested that interpersonal style can indeed add to the predictive validity of violence risk assessment, at least in institutional settings. A study that used the CIRCLE to evaluate forensic in-patients (predominantly with psychotic illnesses) found that dominant, hostile and coercive interpersonal styles were associated with a higher risk of violent behaviour over the following 3 months, whereas a compliant interpersonal style was protective against subsequent violence (Reference Doyle and DolanDoyle 2006). Another forensic in-patient study using the CIRCLE (Reference Daffern, Tonkin and HowellsDaffern 2010a) found that a coercive interpersonal style was associated with increased risk of aggression and self-harm during the following 6 months. A study of patients admitted to acute psychiatric units found that a hostile, dominant interpersonal style measured using the IMI–C predicted violence during the subsequent in-patient stay more strongly than did either acute psychiatric symptoms or patients’ perception of being coerced (Reference Daffern, Thomas and FergusonDaffern 2010b).

Utilising emotions

The systematic use of a clinician’s own emotions as a data source may represent something of a rapprochement between psycho-dynamic and more mainstream approaches to risk assessment. This does not, however, legitimise a return to a pure unstructured approach. Some psychodynamically oriented critics of systematic, empirically grounded approaches to risk assessment have wrongly seen the emphasis on structure as having a defensive quality that militates against genuine understanding of the human interactions involved (Reference MurphyMurphy 2002; Reference DoctorDoctor 2004). Others, however, have recognised that there is in fact no such conflict and that ‘with experience, clinicians begin to recognise countertransferential responses in certain types of clinical situations and, as the likelihood of acting on them reduces, these responses can be utilised as tools both in helping to make the diagnosis, but also to assist in exploring the meaning of the risk’ (Reference FlewettFlewett 2010).

Reference MeehlMeehl (1954) draws a crucial distinction between the relevant facts, which may include ‘immediate impressionistic clinical judgements’ (in which the intuitive subjective clinical impression is treated as a type of fact – the clinician being ‘a testing instrument of a sort’) and methods of combination; such methods, he asserts, can be ‘actuarial’ even for such emotionally based data. He argues that the value of such facts in predicting outcomes is an empirical question to be answered scientifically: ‘it is still an open question whether the fact that the patient acts hostile or dominant ought to be given the weight that the clinician gives it at arriving at his predictions’ [my italics]. Studies such as those described above are beginning, half a century after Meehl’s writing, to empirically address this question.

Emotions about accuracy of judgements

Clinicians also have an emotional sense of the likely accuracy of their own judgement – a level of confidence in their own estimates. ‘Calibration’ refers to the relationship between a clinician’s sense of confidence and subsequent accuracy of their assessment decisions: if clinicians are highly confident about a violence risk assessment and this turns out to be accurate, but less confident about predictions that turn out to be inaccurate, then they are well calibrated. Research in this area is limited. US studies have shown positive relationships between accuracy and confidence for both unstructured (Reference McNiel, Sandberg and BinderMcNiel 1998) and structured (Reference Douglas and OgloffDouglas 2003) professional judgement approaches. However, other studies, such as that by Reference De Vogel and De RuiterDe Vogel & De Ruiter (2004) in the Netherlands, have found that treating clinicians may be overconfident and less accurate in risk assessments involving their own patients. A Canadian study using the Short-Term Assessment of Risk and Treatability (START; Reference Webster, Martin and BrinkWebster 2004) – a structured professional judgement tool – failed to show significant correlations between confidence and accuracy with respect to violence risk assessment (Reference Desmarais, Nicholls and ReadDesmarais 2010). This will be an important area for further research in risk assessment.

Guideline 3: Impose ‘circuit breakers’

Hogarth asserts that the potentially deleterious effect of heuristic biases can be minimised by explicitly and deliberately using ‘circuit breakers’. By this, he refers to explicit processes that stop automatic decisional processing in its tracks and encourage the decision maker to incorporate deliberative reasoning. One such circuit breaker is the well-established process of ‘cost–benefit analysis’: a process that compares options by systematically comparing their possible costs and benefits to determine which is optimal.

When assessing and managing clinical risk, commonly there is an array of competing factors to be considered in this way (Reference Miller, Tabakin and SchimmelMiller 2000), such as:

1. • How will this look if the worst happens?

2. • What are the likely outcomes in the longer term as opposed to the short term?

3. • What is the scientific evidence base and how similar is this patient to patients in that evidence base?

4. • What resources are available?

5. • What are the legal and ethical issues involved?

It is immediately apparent that these competing considerations cover a range of different domains, including the clinical, legal, ethical and political. It has been proposed (Reference Montague and BernsMontague 2002; Reference Peters, Vastfjall and GarlingPeters 2006) that the human decision-making system, when considering pros and cons from such a panoply of different domains, uses affect as a kind of ‘common currency’. This allows us to compare the values of very different decision options: when comparing ‘apples with oranges’, each is first translated to a certain number of ‘affective units’ that may be of positive or negative valence. Reference Peters, Vastfjall and GarlingPeters (2006) describes this process as follows:

‘by translating more complex thoughts into simpler affective evaluations, decision makers can compare and integrate good and bad feelings rather than attempt to make sense out of a multitude of conflicting logical reasons. This function is thus an extension of the affect-as-information function into more complex decisions that require integration of information’.

Hence, the cost–benefit matrix (see Reference CarrollCarroll 2009), a deceptively simple tool, represents an integration of two approaches: the deliberative (systematically ensuring consideration of all logically relevant data) and the intuitive (facilitating the translation of the informational matrix into an overall affective output, favouring one decision pathway over another).

As an example, Table 2 shows a simplified cost–benefit matrix for the clinical dilemma of whether to compulsorily admit to hospital a man with schizophrenia who has disclosed the emergence of violent command hallucinations. A structured, evidence-based approach is used to ensure that all relevant factors are considered. For the purposes of illustration, each consideration has been given an arbitrary score in ‘affective units’. In this example, the clinician’s overall ‘feel’ is in favour of admission to hospital.

TABLE 2 Simplified cost–benefit matrix for deciding whether to compulsorily admit a man with schizophrenia

Violence risk assessment tools

Another circuit breaker is the violence risk assessment tool. The superiority of such tools over unstructured, purely intuitive clinical approaches has now been comprehensively demonstrated (Reference Douglas, Reeves, Otto and DouglasDouglas 2010b). There are, no doubt, various reasons for this, not least that they force the clinician to focus on items of predictive validity and ignore invalid factors. However, they are generally seen not as wholesale replacements for clinical judgement, but as decision support tools (Reference Monahan, Steadman and SilverMonahan 2001). Risk assessment tools can therefore be appropriately used in conjunction with cost–benefit analyses. The complexity of real-world risk management decisions means that simply allocating patients to a category of low, medium or high risk is generally only the start of the decisional task and rarely in itself will be determinative: the pros and cons of the various management options (such as coercive psychiatric care) still need to be systematically considered. This nuanced approach to risk assessment and management is now recognised in the more sophisticated critiques of the state of the field (Reference MossmanMossman 2006), but is unfortunately absent from some polemics (Reference Ryan, Nielssen and PatonRyan 2010).

MCQs

Select the single best option for each question stem

1. 1 Epstein describes the intuitive mode of thinking as:

   1. a rational

   2. b slow

   3. c resistant to change

   4. d more highly differentiated nuanced thinking

   5. e abstract.

2. 2 ‘Kind’ learning contexts have the following characteristics:

   1. a smaller sample sizes

   2. b outcomes that are tightly linked to the judgemental decisions

   3. c delayed feedback

   4. d unrepresentative samples

   5. e low calibration.

3. 3 A dominant interpersonal style on the part of a patient:

   1. a is related to a lower risk of violent behaviour

   2. b elicits corresponding feelings of dominance in the clinician

   3. c can be measured using the Short-Term Assessment of Risk and Treatability (START)

   4. d is a measure of affiliation

   5. e may indicate a specific ‘risk signature’.

4. 4 Structured professional judgement approaches to violence risk assessment:

   1. a attempt to negate the role of clinical judgement

   2. b have been shown to have good reliability and validity

   3. c make cost–benefit analyses superfluous

   4. d always require the Historical-Clinical-Risk Management–20 (HCR–20) to be completed

   5. e indicate whether coercive treatment is indicated.

5. 5 When carrying out a violence risk assessment, countertransference:

   1. a is impossible to test empirically

   2. b should be ignored, to assure maximum reliability and validity

   3. c may be a useful source of data

   4. d is reliably measured by the HCR–20

   5. e should be quickly acted on by the clinician.

MCQ answers

1

c

2

b

3

e

4

b

5

c