Content uploaded by Joachim Diederich
Author content
All content in this area was uploaded by Joachim Diederich on Oct 26, 2018
Content may be subject to copyright.
Content uploaded by Joachim Diederich
Author content
All content in this area was uploaded by Joachim Diederich on Oct 05, 2018
Content may be subject to copyright.
1
Psychotherapy as Heuristic Search
Joachim Diederich
Psychology Network Pty Ltd
Level 1, 16 McDougall St
Milton Qld 4064 Australia
joachim@psychologynetwork.com.au
School of Information Technology and Electrical Engineering
The University of Queensland
Brisbane Qld 4072 Australia
Abstract
There is no coherent framework for psychological therapy available. From psychoanalysis to
behaviour therapy, humanistic and third wave approaches, there are significant differences in
terms of aetiology, methods and expected outcomes. Heuristic search, an important method in
artificial intelligence (AI), can serve as a general frame of reference for psychotherapy. This
approach is based on a clear definition of the starting and end points of psychological therapy
and how to achieve identified outcomes; it allows the utilisation of particularly effective
psychological interventions and a ranking of different forms of psychotherapy with regard to
efficacy, time and effort. A cornerstone of this framework is the notion of "deep heuristic", that
is, particularly effective psychological interventions based on clinical experience. It is likely that
the heuristic search framework can be applied to interventions in psychiatry as well as health
disciplines in general.
1. Introduction
Heuristic search is a form of problem solving. There is a clearly identified start state (e.g. a client
with a set of symptoms) and a goal state (e.g. the client is free of symptoms). Often there are
multiple or complex goals (e.g. client is free of symptoms and well supported by a social
network). There are also means of transforming one state to another: psychological interventions.
These interventions are informed by evidence-based research but clinical practice is important as
well. A significant part of the training of clinical psychologists is based on clinical practice. An
experienced clinician develops heuristic knowledge: facts, skills and methods that are
particularly useful. Sometimes this knowledge is explicit, for instance, when the therapists
knows that s/he can initiate a certain action; sometimes the knowledge is implicit, that is, there is
no conscious awareness of it. Here, deep heuristics refer to psychological interventions that are
deemed particularly useful based on clinical experience in a given context.
Psychological interventions are the means of transforming one state (set of symptoms, level of
well being) to another in order to achieve a clearly defined goal. Psychological interventions can
be ranked at any point in time with regard to a number of criteria: (1) Legal requirements and the
management of risks, (2) empirical support for the intervention, (3) probability of success and (4)
2
confidence of the practitioner to conduct the intervention based on training and clinical
experience. Clearly, the ranking can include other criteria as well.
In summary, artificial intelligence developed the idea that problem solving is search. From this
viewpoint, psychotherapy is the search for means to achieve goal states; freedom of symptoms
frequently among them. The solution to a problem is a viable path from a starting point (the
client presents with symptoms) to one or several goals states (absence of symptoms, general well
being, social support etc).
2. Motivation
The objective here is to offer a unifying approach to psychological therapy. The framework
introduced here facilitates a comparison of different forms of psychotherapy that do not share
common ground with regard to aetiology and the nature of psychological interventions.
Furthermore, the framework considers and integrates knowledge obtained from empirical studies
as well as from clinical practice. A new approach for the measurement of efficacy is introduced
that acknowledges the diversity of psychological interventions and yet focuses on observable and
measurable outcomes. A component of this framework is the preregistration of psychological
treatment similar to empirical, psychological research. The clinician is asked to nominate a set of
assessment methods, psychological interventions and intended outcomes and the result of
psychological treatment is then evaluated in the context of that registration.
Traditionally, problem-solving therapy includes adaptive problem-solving skills to cope with a
stressful problem (American Psychological Association, Division 12, 2018). However, it is
possible to go much further and see problem solving as a general framework for psychological
therapy.
While psychological assessment, and in particular psychometric testing, has a firm theoretical,
mathematical basis, no such coherent and unifying framework is available for psychological
therapy. Heuristic search can serve as a general framework for psychotherapy. A heuristic is a
“rule of thumb”, a method that might not always find the best solution but is guaranteed to find a
good solution in reasonable time. Heuristic search is a form of problem solving that aims at
finding the best and most cost-effective path from a starting point to a well-defined goal. This
approach is based on a clear definition of the starting and end points of psychological therapy
and how to achieve identified outcomes; it allows the utilisation of particularly effective
psychological interventions and a ranking of different forms of psychotherapy with regard to
time and effort.
Daily clinical practice includes the observation of the behaviour of a client, partners, care-givers
and families. It includes the understanding and appropriate response to facial expressions,
gestures, the use of metaphors and sometimes slang language. Much of this knowledge is
embodied in the sense that clients use their full presence to communicate and the psychologist is
using speech, voice, gesture etc. to ask questions and to conduct psychological interventions.
As a result, the “knowledge base” of a clinical psychologist is not only a large network of facts
but "a large array of informal judgement rules (heuristics) which guide the system (here: the
clinical psychologist) toward plausible paths to follow and away from implausible ones (Lenat,
3
1982, p.189). This path includes the individual steps of psychological assessment and therapy.
Furthermore, this heuristic knowledge cannot be directly observed, however, it frequently is the
foundation of daily, clinical practice.
For instance, an experienced psychologists may use “goal decomposition” as a strategy within
the general context of problem-solving therapy. Based on clinical experience and an
understanding of the problem, the psychologist may suggest sub-goals in order to achieve the
overall outcome (the goal). For instance, if the psychologist has worked with clients who are on
the autistic spectrum, the practitioner may may well have a set of heuristics to develop social
communication that can help to achieve sub-goals as well as the final outcome.
3. Case Studies
In order to illustrate the utility of the framework, a number of case studies are summarised below
and the process of psychotherapy is translated into the conceptual structure introduced here.
3.1.1 Historic Case Studies
3.1 Little Albert
Little Albert is among the most famous case studies in psychology. Watson & Rayner (1920)
used classical conditioning to demonstrate how a form of “neurosis” can develop in a child (Trull
& Prinstein, 2013). A young boy who was not afraid of any furry animals was presented with a
white rat and while he was playing with the animal a loud noise was made behind him. After few
trials, Albert developed a fearful response and this generalised to other furry objects (the ethics
of this study and the experimental set-up have been severely criticised for decades).
Jones (1924; in Trull & Prinstein, 2013) demonstrated how fear can be removed in an
independent study with a three-year old boy. Peter was afraid of rabbits and similar animals. In
order to treat this form of anxiety, Jones (1924) brought a caged rabbit closer and closer to the
boy as he was eating. As a result, the dreaded object became associated with food and after some
period of time, Peter’s fear of rabbits disappeared. It was important that Jones (1924) assessed
the fear of rabbits to make sure it was not so intense that the child would develop an aversion to
food. This tightly controlled treatment was successful.
Jones (1924) had a clearly defined starting point for her psychological therapy: Peter and his fear
of rabbits. In addition, there are methods which are now part of behaviour therapy to eradicate
the fear (in this framework, these are the means to move from one sub-goal to the next). The
starting point was Peter with a set of symptoms (fear of a rabbit). The goal state was the client
free of any symptoms. Step-by-step, Jones (1924) brought the cage with the rabbit closer and
closer to Peter. By doing so, she moved from the start state to intermediate states and finally to
the goal state (a client relieved of this particular fear). Please note that while the individual
actions (moving the cage) are almost identical, the intermediate problem solving states in this
particular form of therapy differ. Each state has a different probability to develop an aversion to
food! Assessment was required to make sure that this food aversion did not happen. Clearly, an
intermediate state with a higher probability of an aversion to food is less desirable than a state
that brings the client closer to a freedom of symptoms with low probability of developing
4
negative side effects. Peter’s treatment can be clearly defined with the problem solving
framework.
3.1.2 Little Hans
This very famous case study by Sigmund Freud is about childhood anxiety as well. Hans’ (not
his real name) father belonged to Freud’s circle of followers in Vienna and he noted that Hans
developed a fear of horses. The father wrote to Freud: “No doubt the ground was prepared by
sexual over-excitation due to his mother's tenderness; but I am not able to specify the actual
exciting cause. He is afraid a horse will bite him in the street, and this fear seems somehow to be
connected with his having been frightened by a large penis” (Strachey, 1909). Alternative
explanations for the fear of horses have been offered, including the observation that five-year old
Hans may have been frightened by a collapsing horse on the street.
Freud had a sexual interpretation of the fear of horses broadly based on the concept of the
Oedipus complex. With this understanding Freud started the process of psychoanalysis. There
was no follow-up, however, fourteen years later Hans reported to be fine and well-adjusted
without memory of his childhood phobia.
Freud stated that psychoanalysis is a therapeutic measure and not a scientific investigation
(Wolpe & Rachman, 1960, p.145). It appears that that psychoanalysis, and in particular the
interpretations, provided the means to achieve goals such as freedom of symptoms. In this case,
no fear of horses.
3.2 A fictional, contemporary case study: Maryam
Maryam was referred by her mother Hibba as she is exhibiting anger and conduct problems at
home. Hibba perceived Maryam to be a challenging child since birth. At the age of 2, parenting
advice was sought from a psychologist since Hibba was overwhelmed by the experience of
having a child and this resulted in arguments with her partner Omar as well. Both parents found
it difficult to respond positively to Maryam and aversive interaction patterns emerged, between
both parents we well as the parent-child system. The aversive behaviour of her parents provided
modelling opportunities for Maryam and her self-esteem deteriorated in recent years since her
academic performance is below average. There are significant protective factors: Maryam lives
in a stable home environment; she is physically healthy, both parents are employed and have
sought help for Maryam in the past. In addition, no conduct problems at school have been
reported. Maryam’s behaviour should improve if the family commits to a psychological
treatment based on cognitive behaviour therapy (CBT) and improvements in the family system.
The following long-term goals have been identified:
1. To reduce Maryam’s anxiety, in particular separation anxiety, as well as feelings of panic
due to separation fears.
2. To reduce periods of low and irritable mood as well as episodes of anger.
5
The long-term goals will be achieved if there is (1) a significant reduction in anxiety and
depression scores (self- and parent assessment; T-scores below 60), and (2) a reduction in
observer-rated scores for mood problems. This goal should be achieved after 4-6 months (8-12
sessions).
In addition, the following mid-term goals have been identified: To increased periods away from
mother Hibba without expressed fear or anger as well as sleeping in Maryam’s own bedroom on
the other side of the house. The mid-term goal will be achieved if (1) Maryam attends day-time
play-dates on her own twice a week and (2) sleeps in her own bedroom several times a week
(after 4-8 weeks). Progress towards these goals will be monitored during the first 2-8 sessions
(every session).
Finally, here are the short-term goals for therapy: Maryam to identify negative thinking patterns
related to separation anxiety and replace these with more positive thoughts. This will be
measured by Maryam expressing these skills during sessions and completing homework
assignments for two consecutive sessions. Maryam should achieve these goals within 2-4 weeks
(to be monitored every session).
After identifying clear goals for psychotherapy, the means for transitioning from one state to
another are identified. In this case, CBT and family therapy. Barrett et al. (2001, p.135/6)
observe that CBT alone as well as CBT plus family management/treatment showed greater
improvement on a variety of measures at a 12-month follow-up when compared with a group
without treatment. More specifically, “clients that received family training also showed
significantly greater improvement than the CBT-only group on a number of measures” (Barrett
et al., 2001, p.136).
Byng-Hall (1995, p. 6) argues that a family therapist can provide a “temporarily secure base for
the whole family during therapy”. The overall aim is to “use this secure base to help the family
explore ways of improving the security of the family's attachment network” (Byng- Hall, 1995,
p.7). The family can then use the secure base to establish closer and more satisfying relationships.
After the identification of goals and sets of psychological interventions that allow to move from
one intermediate state to another, the sessions can proceed as follows:
a) First session (with Maryam and both parents): Learning to beat anxiety. The foundations
of CBT are explained to Maryam and Hibba.
b) Second session (with Maryam): Introduction of the “Hot Diary” (a booklet with negative,
problematic thoughts). Several sample situations are discussed.
c) Third session (with the entire family: Explain attachment theory and “circle of security”.
Also, explain the importance of a “secure family base”.
d) Fourth session (with Maryam). Introduction to controlled breathing exercises. Discussion
of the Hot Diary and fears in relation to sleeping in Maryam’s own bed.
In this case study, goal decomposition is used to determine the desired final outcome but also to
identify sub goals that constitute progress in psychotherapy. For benchmarking purposes, it is
clearly stated which goals should have been reached after a certain number of sessions.
Interventions from cognitive behaviour therapy as well as family therapy are the actions that
6
allow to traverse problem space, that is to move from one sub-goal to the next until the final
outcome is achieved. In addition, the criteria for achieving goal state are very clear: scores in
psychological tests that measure child anxiety.
4. Knowledge acquisition and automated psychotherapy: Kriton Speech
Kriton Speech integrates cognitive computing and psychological methods to acquire knowledge
for artificial intelligence systems and to apply this knowledge for computational diagnosis and
therapy. Kriton was developed as a general knowledge acquisition system and has been adopted
for psychological applications (Diederich 1987a,d, 1988). As a knowledge acquisition system,
Kriton Speech elicits domain specific knowledge and as an application, it applies this knowledge
for diagnosis and therapy. Kriton Speech exclusively uses a voice user interface (VUI). The
system uses knowledge bases such as upper-level ontologies as well as domain-specific
knowledge and psychological heuristics (1) to interview the user to obtain knowledge and based
on this knowledge (2) perform tasks such as psychological assessment and therapy. The system
allows free verbal responses and has an explanation capability, i.e. provides feedback about why
certain questions are asked. Speech and language are analysed to determine, with high accuracy,
if there is a mental health problem.
Attention economy is a concept in information management that treats human attention as a
limited commodity. Herbert A. Simon (1971) is considered to be the first person having
introduced the concept of attention in economics. Simon (1971) outlined that in an information
rich world there is a limitation of the resource that consumes information: human attention.
Almost half a century later, we are now at the point where there are so many information sources,
so many devices and so many distractions that human attention is not only limited, it is
frequently impaired.
Kriton Speech, an artificial intelligence system, addresses both problems: the limitation of
human attention and the challenge of growing knowledge. Firstly, Kriton Speech uses speech
only to interact with the user and therefore does not require any visual attention. This enables the
user to direct attention to other information sources and to perform physical actions while using
the program. No learning is required, any user can interact with Kriton Speech immediately. It is
possible to drive a car while talking to Kriton Speech; to go for a walk or to perform other
activities.
Kriton Speech is using upper-level ontologies as well as existing domain-specific knowledge to
interview the user. The employment and use of already acquired knowledge has major
advantages: (1) The interview aims at completing and refining existing knowledge in addition to
the acquisition of new concepts and rules. (2) On the application side, the system utilises up-to-
date clinical knowledge obtained from practising mental health experts to conduct psychological
assessment and therapy.
As a result of the interview, Kriton Speech builds and verifies ontologies and rule based systems.
The output is a Web Ontology Language (OWL) file that can be edited by the use of standard
ontology editors. Information obtained from clients is stored in an XML format to be used by
eHealth systems.
7
Figure 1: An ontology for ADHD coaching.
4.1 Amy: An artificial intelligence ADHD coach
Amy is an artificial intelligence (AI) system that provides psychological assistance simply by
talking to a client. Amy is available 24 hours a day and provides an additional dimension for
psychological service delivery. When linked to background information in the form of ontologies,
Amy has extensive medical and psychological knowledge and the built-in capacity to acquire
additional knowledge from mental health experts. In turn, this knowledge is used for psycho-
education. Amy’s primary mode of communication is a simple and convenient conversation,
however, Amy also provides videos, images and text to educate users about mental health issues.
Amy operates in a number of different modes. If the user is inclined to have a free conversation
about any topic or to obtain information about ADHD, Amy is simply a chatbot with a very
strong protection against inappropriate content. With machine learning enabled, Amy detects
mood changes in speech and text including those pointing to depression and can modify the
interaction accordingly. Longer conversations can be analysed to extract personality
characteristics of the user. While Amy has the ability to learn, the dialogue structure is rule-
based and does not change. As a result, Amy is always socially and culturally appropriate. Amy
frequently refers to online resources to illustrate recommendations, including the use of
relaxation techniques.
4.1.1 Technical Solution.
Many of the current "artificial psychologist" solutions go wide, i.e. they aim at providing
counselling on almost any subject. As a result, these solutions do not have in-depth knowledge
8
about mental health disorders and they cannot provide therapy that is specifically tailored to
address complex psychological conditions. The approach here is narrow: Amy uses extensive
medical and psychological knowledge bases, including knowledge provided by practising
clinical psychologists. All these knowledge sources are used to perform tasks such as psycho-
education and coaching but only in the context of ADHD! Amy’s knowledge bases are updated
frequently to include the latest facts about mental health and ADHD. Currently, Amy is
interacting in English language. Multi-lingual implementations in a number of Asian language
are expected soon. Amy interacts primarily through social media (e.g. Twitter).
Figure 2: A generic ontology for clinical psychology.
5. Conclusion
The framework introduced here applies to all forms of psychological service deliver, from the
clinical practise of trained psychologists to computational approaches. Heuristic search can serve
as a general frame of reference for psychotherapy. This approach is based on a clear definition of
the starting and end points of psychological therapy and how to achieve identified outcomes; its
usefulness is demonstrated by use of case studies. It is likely that the heuristic search framework
can be applied to interventions in psychiatry as well as health disciplines in general. The
approach allows the utilisation of particularly effective psychological interventions and a ranking
of different forms of psychotherapy with regard to time and effort.
References
Barrett, P. M., Duffy, A. L., Dadds, M. R., & Rapee, R. M. (2001). Cognitive-behavioral
treatment of anxiety disorders in children: long-term (6-year) follow-up. Journal of Consulting
and Clinical Psychology, 69, 135–141.
9
Diederich, J., Techniken des Wissenserwerbs (Knowledge Acquisition Techniques), in
Christaller, Th. (Ed.), 5. Künstliche Intelligenz Frühjahrsschule (1987). Berlin: Springer Verlag
1988 = Springer Informatik Fachberichte 202, 295–335.
Diederich, J., Ruhmann, I., May, M., KRITON: A Knowledge Acquisition Tool For Expert
Systems. International Journal of Man-Machine Studies 26 (1987) 29-40.
Diederich, J., Ruhmann, I., May, M., KRITON: A Knowledge Acquisition Tool for Expert
Systems, in Boose, J.H., Gaines, B. (Eds.), Knowledge Acquisition Tools for Expert Systems.
Knowledge-Based Systems 2. London etc: Academic Press (1988) 83–94.
Lenat, D.B., The Nature of Heuristics. Artificial Intelligence, 19 (1982) 189-249.
Strachey, J. Analysis of a Phobia in a Five-Year-Old Boy. The Standard Edition of the Complete
Psychological Works of Sigmund Freud, Volume X (1909): Two Case Histories (‘Little Hans’
and the ‘Rat Man’), 1-150.
Sijbrandija, M., Acarturkb, C., Birdc, M., Bryantd, R.A., Burcherte, S. ……. Strengthening
mental health care systems for Syrian refugees in Europe and the Middle East: integrating
scalable psychological interventions in eight countries. European Journal of Psychotraumatology,
2017, VOL. 8, 1388102 https://doi.org/10.1080/20008198.2017.1388102.
Simon, H. A. (1971) "Designing Organizations for an Information-Rich World" in: Martin
Greenberger, Computers, Communication, and the Public Interest, Baltimore. MD: The Johns
Hopkins Press. pp. 40–41.
Trull, T.J., Prinstein, M.J. Clinical Psychology (Eighth Edition). Wadsworth CENGAGE
Learning, Belmond CA, 2013.
What is Problem-Solving Therapy? American Psychological Association | Division 12
http://www.div12.org/sites/default/files/WhatIsProblemSolvingTherapy.pdf Retrieved 22 March
2018.
Wolpe, J., Rachman, S. Psychoanalytic "evidence": A critique based on Freud's case of little
Hans. Journal of Nervous and Mental Disease, 131 (1960) 135-148.
