Kisan Call Centre - To harness potential of ICT in Agriculture by answer farm...
Charting the explanatory potential of network models/network modeling in psychopathology
1. Charting the explanatory potential of network
models/network modeling in psychopathology
Federica Russo & Dingmar van Eck
University of Amsterdam
Department of Philosophy & Institute for Logic, Language & Computation
2. Aim
• Charting the explanatory potential of network models/network
modeling in psychopathology
• Running case: network model of panic disorder (Robinaugh et al. 2019)
2
3. Motivation
• Network theorists assert that the explanation of mental disorders by reference to
underlying common causes will by and large not succeed and propose an
alternative explanatory strategy (Borsboom et al. 2019).
• We submit: the current methodological literature is underdeveloped:
• What does this alternative explanatory strategy precisely entail?
• What do network theorists exactly have in mind when speaking about ‘explanation’?
• (See also de Boer et al. 2021).
• We offer an account that clarifies what explanatory alternative to common cause
explanations is exactly being offered by the network approach.
3
4. Our approach in a nutshell:
• Explanatory potential
• Network models as qualified mechanistic models
• Network models track difference making relations at the mechanistic network structure
level
• Heuristics
• Network models as heuristics for hypothesis formulation and testing
• System demarcation
• Shaping and reshaping fluid system boundaries of network models
• The more a network model is able to adapt and shape explanations across different
configurations, the higher its explanatory power.
4
5. Outlook
• Brief recap on network theory of mental disorders
• Our account of explanation in the network approach:
• Explanatory potential
• Heuristics
• System demarcation
5
6. Network theory of mental disorders
• The network theory of mental disorders conceptualizes mental disorders in terms
of networks of causally connected symptoms
• Mental disorders: stable, dysfunctional states in which such networks can get
locked
• Alternative to latent variable or common cause models of mental disorders in
which symptoms are understood as arising from (and indicators of) underlying
common causes, viz. common disease mechanisms or pathogenic pathways
• According to advocates of the network approach, very few (if any) of such
common disease mechanisms for mental disorders exist
6
8. In the words of network theorists:
“For instance, if one thinks that other people can read one’s mind (delusion), this may generate
extreme suspicion (paranoia); this paranoia can lead one to avoid other people (social isolation),
which, because one is no longer exposed to corrective actions of the social environment, may serve
to sustain and exacerbate the relevant delusions. In this way, symptoms may form feedback loops
that lead the person to spiral down into the state of prolonged symptom activation that we
phenomenologically recognize as a mental disorder.” (Borsboom 2017, pp 5-6).
9
9. Principles of the network theory of mental
disorders
5 principles of the network theory of mental disorders
Complexity
Interaction between different components in a network
Symptom-component correspondence
Components are symptoms
Direct causal connections
Network structures are patterns of causal connections between symptoms
Mental disorders follow network structure
Symptoms are causally connected/grouped in specific ways
Hysteresis
Networks are self sustaining: causal connections persist when the triggers have vanished
10
10. Explanatory potential
• Network models are qualified mechanistic models
• We work with ‘Minimal Mechanism’:
• “a mechanism for a phenomenon consists of entities (or parts) whose activities and interactions are
organized so as to be responsible for the phenomenon” (Glennan et al., 2022, p 145).
• Key elements of mechanisms and mechanistic models thereof:
• (descriptions of) activities,
• interactions,
• causal and temporal organization
• (constitution).
12
11. What exactly do we explain?
• Network models of mental disorders afford explaining the covariation between two or more
symptoms by clarifying
a) when a symptom is (and when it isn’t) a difference maker for the occurrence of another symptom, and
b) succeed in explaining variations in the strength of such causal symptom-symptom relations
• But, importantly, to assess these changes adequately, i.e., to understand the occurrence and
persistence of mental disorders and their changing nature, one needs to take into account how
these changes affect the whole network structure, i.e., impact the mental disorder as a whole.
13
13. Network models are heuristic tools
• Network models also enable us to generate hypotheses about which symptom-symptom relations
or configurations of symptoms in the model to further scrutinize and test in individual therapeutic
settings.
• We could use generic network models by inquiring into how patients and clinicians view the
relations articulated in a generic model, possibly identifying new relations and possibly
questioning the relevance of other relations, thereby testing and possibly refining the generic
model.
• In this sense, we claim, network models are (in addition to explanatory models also) heuristic
tools.
• The heuristic perspective goes two ways
• from a generic model to specific individuals (e.g., to partly inform therapy) and
• from specific individuals to a generic model (e.g. to explore other connections or measurement).
• This is in line with mechanistic property cluster view, giving it a methodological orientation
17
14. How to demarcate boundaries of systems
• Often, in the mechanism literature, system boundaries are thought to be rigid/robust
configurations (whence classic understanding of ‘constitutive relevance’).
• We claim: things are exactly the opposite in the case of network models:
• Mental disorders are changing, evolving, and constituted by their own symptoms (and not results of hidden,
common cause, variables).
• Network structures and thus system boundaries are dynamical and prone to change.
• Establishing the robustness of the system boundary -- the configuration of the symptoms -- as a proxy for
explanatory power of the network model runs counter to explanatory purposes in the case of network modeling.
• The more a network model is able to adapt and shape explanations across different
configurations, the higher its explanatory power.
• The point of network modeling is to identify the relevant symptoms, and to clarify how they may change, as
conditions change.
• For instance, see example of panic attack and disorder,
• How a decrease in physiological arousal elicits a decrease in the perception of threat which, in turn, affects escape behavior
which, in turn, may lead to changes in arousal and appraisals thereof, thereby changing the perception of the severity of
threat.
• Given the dynamical, changing nature of network models and hence of system boundaries, attempts at system demarcation by
testing for constitutive relevance, as advertised in the philosophical literature on mechanistic explanation, are not quite viable.
18
15. To sum and conclude
• There is much to learn from and think about network models of
mental disorders
• We find the approach very promising and we hope to contribute to its
development by clarifying what their explanatory potential is really
about.
• In short:
• Network models explain co-variations
• They can be used as heuristic tools to generate novel generic hypotheses or
give directions to individual treatment
• System boundaries have to be fluid, and so there are no ‘strong’ constitutive
relevance relevance relations in these mechanisms
19
16. Charting the explanatory potential of network
models/network modeling in psychopathology
Federica Russo & Dingmar van Eck
University of Amsterdam
Department of Philosophy & Institute for Logic, Language & Computation
Thanks for your attention
Editor's Notes
The motivating problem
Model is a summary of a large body of empirical psychological findings