# Full Text: Aligning Active Inference Ontology to SUMO

> Extracted from `2024_OntologySUMO.pdf`

---

## Page 1

Aligning Active Inference Ontology to SUMO
Authors: David S. Douglass1,2 https://orcid.org/0000-0001-7894-8019; Adam Pease3 https://orcid.org/0000-0001-9772-1266; Daniel Friedman1 https://orcid.org/0000-0001-6232-9096; Jessica 
Balbuena1 https://orcid.org/0000-0003-2741-4476; Rhea Chokhalingam1 https://orcid.org/0009-0007-2061-3022; Ana Magdelena Hurtado4,1 https://orcid.org/0000-0003-4064-1876; Maria Luiza 
Iennaco5,1 https://orcid.org/0000-0002-5407-4852; V. Bleu Knight1 https://orcid.org/0000-0002-9894-1989; Scott Ryan Maybell6,1 https://orcid.org/0000-0003-1612-5505; Ali Rahmjoo1 
https://orcid.org/0000-0002-3244-7419; Paulo Duare Andrade Sayeg5,1 https://orcid.org/0009-0006-0658-7684; Jakub Smékal7 https://orcid.org/0000-0003-4989-4968; Dean Tickles8,1 
https://orcid.org/0000-0003-2213-0773; Alex Vyatkin1 https://orcid.org/0000-0003-1306-4620
1 Active Inference Institute
2 American Society for Cybernetics
3 Naval Postgraduate School
4 Arizona State University
5 University of São Paulo
6 University of Oxford
7 Stanford University
8 Professional Initiatives Programming
DOI: 10.5281/zenodo.11459323
Contact: DavidSDouglass@ActiveInference.Institute
ABSTRACT
Active Inference and its associated concepts (notably predictive coding, predictive processing, and the Free Energy Principle) act instrumentally as a conceptual system that provides structure to a 
large and growing number of scientific and humanistic fields; and realistically as candidate real-world properties and their mutual affordances. 
In 2021, the Active Inference Institute adopted SUMO, the Suggested Upper Merged Ontology, as the ontology with which to map its inventory of topics and their interrelations, along with the 
corresponding multilingual technical terms.
This document aims to act as a scaffolding, displaying a tentative alignment of Active Inference terms with SUMO entities, in order to facilitate the anticipated process of rigorously mapping Active 
Inference to SUMO. For a subset of Active Inference terms, we identify published SUMO files (i.e. .kif files) likely to contain correctly delineated mappings of those topics; likely SUMO supersets of 
the desired targets; some exemplary SUMO subsets of the targets; and indications of relationships among the identified SUMO topics.
KEYWORDS: Active Inference Ontology, SUMO Ontology

## Page 2

OUTLINE OF CONTENTS
1. Notes on Presentation of Active Inference Data.
2. Notes on Presentation of SUMO Data.
3. An “alignment” table. This shows, for many core terms used in the Active Inference Ontology, areas of SUMO that are likely to contain corresponding topics; or which are portions of SUMO where 
topics should be created to more nearly match Active Inference concepts.
1. Notes on Presentation of Active Inference Data.
Some of the following items occur only in the external full table of alignments of Active Inference to SUMO, i.e. “Active Inference Topics Aligned to SUMO.xlsx”
Concept. A concept is a persisting process of reproducing (copying, rehearsing, …) a relation, either perfectly or with constrained mutation. See Pask and Scott (1973).
Topic. A topic is a shared concept. Distinguished from term.
Term. A term is a linguistic artifact, written or spoken, in a specific natural or formal language, used to indicate (activate, represent, …) a topic.
Module. Active Inference topics are grouped into a small number of modules (also called sub-theories or subontologies). These are high-level clusters of topics that show a high degree of inter-
definition, interdependence, and collocation in texts. In a typical course, members of a module would taught together.
As of 2024, the inventory of modules in the Active Inference ontology is rather informal. The module, with some of their notable members, are:
Action Action Planning, Action Prediction, Agency, Behavior, Policy, Policy selection, Preference
Agents in the Niche Affordance, Agent, Cognition, Culture, Ensemble, Generative Model, Generative Process, Narrative, Niche, Non-Equilibrium Steady State, Particle, Recognition Model… 
Bayesian Statistics Accuracy, Ambiguity, Bayes Theorem, Bayesian Inference, Belief, Belief updating, Complexity, Data, Expectation, Inference, Information, Learning, Outcome, Posterior…
Free Energy Active Inference, Epistemic value, Expected Free Energy, Free Energy Principle, Generalized Free Energy, Pragmatic Value, Process Theory, Variational, Variational Free Energy
Information Cue, Information Geometry
Markov Partitioning Active States, Blanket State, External State, Friston Blanket, Hidden State, Internal State, Markov Blanket, Markov Decision Process, Sense State
Perception Attention, Evidence, Novelty, Observation, Regime of Attention, Salience
Systems 
Hierarchical Model, Living system, Multi-scale system
List. Each individual topics has been assigned to one of three “lists.” This is a porous partition, created to help organize exposition.
Core topics should be mastered in a general course in Active Inference concepts, aimed at upper-division college undergraduates in technical curricula. 
A textbook might spend from half a page to one section of a chapter, on each core topic. 
Omitting or failing to master a core topic can be regarded as a failure of presentation or learning.
Each entailed topic is essential to understanding at least one core topics. An instructor or textbook author can assume that some of these topics are already understood by a typical adequately-
prepared student. Nonetheless, usage of each entailed topic should be indicated, i.e. “defined,” in at least as much detail as is typically found in a technical glossary in a textbook or review 
article. Source: Definitions of core terms.

## Page 3

Supplemental topics define notions typical of “extra credit” or “advanced topics” material. As conceived by the Active Inference Ontology working group, adding a supplemental topic to a 
presentation expands the scope of ideas covered, but need not be considered essential to mastery of Active Inference and its major cohorts (notably Predictive Process and the Free Energy 
Principle).
Definitions. We restrict ourselves to two senses or three senses of a given topic name. Deeper examination may show that some of these “senses” should be separated into distinct topics (which may 
be analogs of one another).
Expositions. The best single source for understanding Active Inference is textbook “Active Inference: The Free Energy Principle in Mind, Brain, and Behavior" by Thomas Parr, Giovanni Pezzulo, 
and Karl J. Friston (2022, MIT Press). This work is accessible online at https://direct.mit.edu/books/oa-monograph/5299/Active-InferenceThe-Free-Energy-Principle-in-Mind
A suitable instructor-run course might run for half a year. An example is the online course “Active Inference Textbook Group,” anchored in Parr, Pezullo, Friston.
Two presentations of this course are viewable at 
https://www.youtube.com/watch?v=G9GfOMjF4g0&list=PLNm0u2n1Iwdob1pSM1q9yzDboE3uvQqq- and
https://www.youtube.com/watch?v=3U8AXcIaUFI&list=PLNm0u2n1Iwdpm1wcq9DOGSdKDDvnEt_xG

## Page 4

2. Notes on Presentation of SUMO Data.
(1) General remarks. A 2001 overview of SUMO can be found at https://www.adampease.org/FOIS.pdf
Unlike taxonomies or knowledge graphs, SUMO concepts are defined axiomatically in SUMO.  The identifiers mean only what their formal axioms say that they mean in mathematical logic, without 
recourse to human interpretation.  The existing set of 100,000 formal statements can be added to in order to define additional terms specific to any domain.
A playlists of Adam Pease's videos, discussing "how to do SUMO," appears at
https://www.youtube.com/watch?v=SkruxPmN0kk&list=PLpBQIgki3izcrbaiuOH_OWWguY_duSumA&ab_channel=OntologyTalkwithAdamPease
(2) Class and relation names. SUMO identifiers use the convention of CamelCase.  Relations that are not functions have an initial lowercase letter.  All other identifiers begin with an uppercase letter.
Sumo variable names are prefixed with “?”.
In this document, we often flag terms that do not yet belong to SUMO by prefixing an asterisk (as in “*ActiveState”).
(3) Column “SUMO superclass, instanceOf, superrelation.” In the tables aligning Active Inference to SUMO terms (both the full table presented separately and the except below), this column 
situates a conceptual entity in the whole SUMO ontology by showing the term’s place one or more “upward” or “including” or contextual or more-general relationships:
superclass. (A neologism introduced in this present document.) See the inverse of superclass, i.e. “subclass,” below.
instanceOf. SUMO describes the instance relation as follows: "An object is an instance of a Class if it is included in that Class. An individual may be an instance of many classes, some of 
which may be subclasses of others. Thus, there is no assumption in the meaning of instance about specificity or uniqueness.")
superrelation. (A neologism introduced in this present document.) See the inverse of superrelation, i.e. “subrelation,” below.
(4) Column “SUMO subclasses, subrelations.” This refines the reader’s appreciation of a SUMO entity name, by giving terms that are “included-in” or less-general than the named target entity:
Subclass. The subclass relation resembles the word “subset” (as used commonparlance to show relations among concepts, e.g. by means of Venn diagrams), but there is no arbitrariness, nor 
exceptions to inheritance: The attributes and relations of a superclasses are inherited by their subclasses.
SUMO describes the subclass relation as follows: "’ CLASS1 is a subclass of CLASS2’ means that every instance of CLASS1 is also an instance of CLASS2. A Class may have 
multiple superclasses and subclasses."
subrelation. 
SUMO describes the subrelation relation as follows: "’REL1 is a subrelation of REL2’ means that ‘every tuple of REL1 is also a tuple of REL2.’ In other words, if the Relation 
REL1 holds for some arguments arg_1, arg_2, ... arg_n, then the Relation REL2 holds for the same arguments. 
A consequence of this is that a Relation and its subrelations must have the same number of arguments.” One further consequence is to distinguish some specfic 
subrelations from relevant metaphorical relations (in which an argument in the target of the metaphor may range freely across possible values without invalidating the 
statement of metaphoricality).

## Page 5

(5) Column “Domains.” This column breaks down a relation that’s relevant to the Active Inference term in question, by going through the related terms, and stating the domain (type, class) of each 
“term of the relation,” from first onward.
SUMO describes the domain relation as follows: "Provides a computationally and heuristically convenient mechanism for declaring the argument types of a given relation. The 
formula (domain REL INT CLASS) means that the INT'th element of each tuple in the relation REL must be an instance of CLASS. Specifying argument types is very 
helpful in maintaining ontologies. Representation systems can use these specifications to classify terms and check integrity constraints. If the restriction on the argument 
type of a Relation is not captured by a Class already defined in the ontology, one can specify a Class compositionally with the functions UnionFn, IntersectionFn, etc.") 
(6) Column “Other SUMO relations.” This column states a variety of relations among SUMO topics and terms. Sometimes the intention is simply to motivate the suggested aligment
with Active Inference.

## Page 6

3. SELECTED ACTIVE INFERENCE TOPICS ALIGNED WITH SUMO CLASSES AND RELATIONS
Topic
Module
Proposed Definition 1
Proposed Definition 2
SUMO superclass, 
instanceOf, superrelation
Domains, Roles
Other SUMO relations
abstractCounte
rpart
 
(abstractCounterpart ?AB
?PHYS) relates a Physical 
entity to an Abstract one 
which is an idealized 
model in some dimension
of the Physical entity. 
Example: an Abstract 
GraphNode could be 
stated to be the 
counterpart of an actual 
Computer in a 
ComputerNetwork.
abstractCounterpart is a 
subrelation of represents. 
abstractCounterpart is an 
instance of binary predicate. 
The number 1 argument 
of abstractCounterpart is 
an instance of abstract. 
The number 2 argument 
of abstractCounterpart is 
an instance of physical. 
 
Accuracy
Bayesian 
Statistics
Broad sense: how “close 
to the mark” an 
Estimator is. 
Narrow sense: the 
expected or realized 
extent of Surprise on an 
estimation, usually about
Sense State reflecting 
the Recognition density 
*Accuracy is a subclass of 
PsychologicalAttribute
An instance of *Accuracy 
is the number 2 argument 
of abstractCounterpart.
*Accuracy is internally 
related to TruthValue
Action
Action
Broad sense: The 
dynamics, mechanisms, 
and measurements of 
Behavior 
Narrow sense: The 
sequence of Active 
States enacted by an 
Agent via Policy selection
from Affordance 
*AbstractAction is a subclass 
of IntentionalProcess.
 
 
Action Planning
Action
The selection of an 
Affordance based upon 
Inference of Expected 
Free Energy 
 
Planning is a subclass of 
IntentionalPsychologicalProc
ess.
 
 
Action 
Prediction
Action
Inference on current and 
future Expectation of 
Action
 
 
 
Predicting is a subclass 
of intentional 
psychological process
Active
 
 
 
actionTendency is an 
instance of binary relation. 
The number 1 argument 
of actionTendency is an 
instance of emotional 
state. 
The number 2 argument 
of actionTendency is a 
subclass of emotional 
behavioral process. 
@active is internally 
related to 
actionTendency. 
Active 
Inference
Free 
Energy
Active Inference is a 
Process Theory related to
Free Energy Principle.
 
Judging is a subclass of 
Selecting 
 
*ActiveInference is 
internally related to 
Judging.
Active States
Markov 
In the Friston Blanket 
 
 
 
*ActiveState is a

## Page 7

Partitionin
g
formalism, the Blanket 
State are the Sense State
(incoming Sensory input) 
and Active States 
(outgoing influence of 
Policy selection)
subclass of 
PhysiologicProcess.
Active Vision
 
refers to the process of 
visual perceptions, in 
terms of oculomotor 
Sensorimotor Behavior 
and Cognitive System 
Generative Model 
refers to regime of visual 
perceptions through 
dynamical perturbations 
of light over the retina. 
As if feeling the textured 
light reflected off the 
niche surfaces
Looking is a subclass of 
intentional process. 
Searching is a subclass of 
investigating.
 
*ActiveVision is a 
subclass of Looking. 
*ActiveVision is a 
subclass of Searching. 
Affordance
Agents in 
the Niche
Options or capacities for 
Action by an Agent 
(sometimes called 
“Affordance 3.0”) 
From Ecological 
Psychology, the 
Perception of a possibility
for Action (sometimes 
called “Affordance 2.0”).
resource is an instance of 
CaseRole.
resource is a subrelation of 
patient.
resource is an instance of 
case role. 
resource is a subrelation 
of patient.
*Affordance is 
equivalent to resource.
Agency
Action
The ability of an Agent to
engage in Action in their 
Niche and enact Goal-
driven selection or Policy 
selection based upon 
Preference 
 
Intentional process is a 
subclass of process
 
*Agency is internally 
related to 
IntentionalProcess
Agent
Agents in 
the Niche
 Entity as modeled by 
Active Inference, with 
Internal State separated 
from External State by 
Blanket State 
 
agent is a subrelation of 
involvedInEvent. 
agent is an instance of case 
role.
The number 1 argument 
of agent is an instance of 
Process. 
The number 2 argument 
of agent is an instance of 
AutonomousAgent.
A SentientAgent is an 
Agent that is capable of 
Perception and 
experiences some level 
of consciousness. 
Ambiguity
Bayesian 
Statistics
Broad sense: Extent to 
which stimuli have 
multiple plausible 
interpretations, requiring 
priors &/or Action for 
disambiguation
Narrow sense: Specific 
model parameter used to
model Uncertainty, 
usually about sensory 
Perception. 
*Ambiguity is a subclass of 
StateOfMind.
 
 
Attention
Perceptio
n
Broad sense: Generative 
Model that is aware of 
some Stimulus, reflected 
by its Salience 
Narrow sense: Attention 
modulates the 
confidence on the 
Precision of Sense State, 
reflecting Sensory input 
*Attention is a subclass of 
IntentionalProcess.
*Attention nominalizes the
attends CaseRole
 
Autopoiesis
 
Phenomena of a System 
that recapitulates the 
 
*Autopoiesis is internally 
related to Relication.

## Page 8

material and 
informational causes of 
its own 
composition/existence. 
Replication is a subclass of 
OrganismProcess.
Bayesian 
Inference
Bayesian 
Statistics
As opposed to frequentist
analysis, Bayesian 
Inference uses a 
specified Prior or 
Empirical prior to Update 
the distributional 
Posterior 
 
*BayesianInference is a 
subclass of 
PhysiologicProcess.
 
 
Behavior
Action
The sequence of Action 
that an Agent is 
observed to enact.
 
 
 
*Behavior is an near 
synonym of BodyMotion.
*Behavior is an near 
synonym of Process.
Belief
Bayesian 
Statistics
Broad sense: Felt sense 
by an Agent of 
something being true, or 
confidence it is the case.
Narrow sense: the State 
of a Random variable in a
Bayesian Inference 
scheme.
*Belief is a subclass of 
PsychologicalProcess. 
Believes is an instance of 
PropositionalAttitude.
 
 
Belief updating
Bayesian 
Statistics
Belief updating is 
changes in a Bayesian 
Inference Belief through 
time. 
 
*BeliefUpdating is a subclass
of 
IntentionalPsychologicalProc
ess. 
IntentionalPsychologicalProc
ess is a subclass of 
IntentionalProcess.
 
 
Blanket State
Markov 
Partitionin
g
Set of states in the 
Markov Blanket Partition 
that make Internal State 
and External State have 
Conditional Probability 
that are independent. 
 
*ThermodynamicBlanketStat
es are PhysicalStates. 
*HomeostaticBlanketStates 
are InternalAttribute.
 
 
Cognition
Agents in 
the Niche
An Agent modifying the 
weights of its Internal 
State for the purpose of 
Action Planning and/or 
Belief updating. 
(This is a 
@realistCounterpart of 
 
CognitiveAgent is a subclass 
of SentientAgent
 
A CognitiveAgent is an 
Agent that has the 
ability to reason, 
deliberate, make plans, 
and experience 
emotions.

## Page 9

Goal-driven selection.) 
Complexity
Bayesian 
Statistics
The extent to which an 
Agent must revise a 
Belief to explain 
incoming Sensory 
observations.
The Kullback-Leibler 
Divergence between the 
Prior and Posterior which 
is used in Bayesian 
model selection to find 
the simplest (least 
complex) model and 
avoid overfitting on the 
noise inherent in Sensory
observations.
*Complexity is a subclass of 
ObjectiveNorm.
 
 
Cue
Informati
on 
A Stimulus, event, object,
or Guidance signal that 
serves to guide Behavior,
such as a retrieval cue, 
or that acts as a @Signal 
to the presentation of 
another stimulus, event, 
or object, such as an 
unconditioned stimulus 
or reinforcement. 
(dictionary.apa.org)
 
 
 
A *Cue is internally 
related to an instance of
Perception. 
AgentPatientProcess is a
subclass of Process.
Culture
Agents in 
the Niche
Culture is the Niche for 
social Agent, that 
structures their Regime 
of Attention 
 
*Culture is a subclass of 
Proposition.
 
 
Data
Bayesian 
Statistics
Data are a set of values 
of qualitative or 
quantitative variables 
about one or more Agent
or object.
 
InformationMeasure is a 
subclass of 
ConstantQuantity. 
Stating is a subclass of 
LinguisticCommunication. 
UnitOfInformation is a 
subclass of 
NonCompositeUnitOfMeasure
The number 1 argument 
of ContainsInformation is 
an instance of 
ContentBearingPhysical. 
*Data is a near synonym
of InformationMeasure.
*Data is a near synonym
of FactualText.
*Data is a near synonym
of Stating. Content 
bearing object is 
internally related to 
contains information.
Content bearing 
physical is a subclass of 
physical. 
Decision-
making
 
Within Active Inference, 
this is the same as Policy
selection 
 
Deciding is a subclass of 
Selecting.

## Page 10

Selecting is a subclass of 
IntentionalPsychologicalProc
ess.
Ensemble
Agents in 
the Niche
Group of more than one 
Agent.
 
*Ensemble is a subclass of 
Collection
 
 
Epistemic value
Free 
Energy
Epistemic value is the 
value of Information gain
or Expectation of 
reduction in Uncertainty 
about a State with 
respect to a Policy, used 
in Policy selection 
 
*EpistemicValue is a 
subclass of 
PsychologicalProcess. 
*EpistemicValue is a 
subclass of 
SubjectiveAssessmentAttribu
te. 
*EpistemicValue is an 
instance of InternalAttribute
 
The abstract 
counterpart of an 
*EpistemicValue is an 
*AbstractEpistemicValue
.
*EpistemicValue is a 
relatedInternalConcept 
to Investigating.
Evidence
Perceptio
n
Data as recognized and 
interpreted by 
Generative Model of 
Agent 
 
*Evidence is internally 
related to 
IntentionalPsychologicalProc
ess.
 
 
Expected Free 
Energy
Free 
Energy
Measure for performing 
Inference on Action over 
a given time horizon 
(Policy selection, Action 
and Planning as 
Divergence 
Minimization). 
The two components of 
Expected Free Energy 
are the imperative to 
satisfy Preferences, and 
the penalty for failing to 
minimize Expectation of 
Surprisal.
*ExpectedFreeEnergy is a 
subclass of 
RelationalAttribute.
 
*ExpectedFreeEnergy is 
internally related to 
InformationMeasure.
External State
Markov 
Partitionin
g
States with #r65 
independent from 
Internal State, 
conditioned on Blanket 
State. 
 
*ExternalState is a subset of 
PhysiologicProcess.
 
 
Free Energy
Free 
Energy
Free Energy is an 
Information Theoretic 
quantity that constitutes 
an upper bound on 
Surprisal. 
Free Energy can refer to 
various or multiple sub-
types of Free Energy: 
* Variational Free Energy 
* Expected Free Energy 
* Free Energy of the 
Expected Future 
* Helmholtz Free Energy 
*FreeEnergy is a subclass of 
PhysicalDimension. 
*FreeEnergy is a subclass of 
RelationalAttribute.
 
*FreeEnergy is internally
related to 
InformationMeasure.

## Page 11

Free Energy 
Principle
Free 
Energy
A generalization of 
Predictive Coding (PC) 
according to which 
organisms minimize an 
upper bound on the 
Entropy of Sensory input 
(or sensory signals) (the 
Free Energy). Under 
specific assumptions, 
Free Energy translates 
toPrediction error. 
A set of statistical 
principles that describe 
how Agents can maintain
their self-organization in 
the face of random 
fluctuations from the 
environment.
*FreeEnergyPrinciple is an 
instance of Proposition.
 
 
Friston Blanket
Markov 
Partitionin
g
Markov Blanket with 
partitioned Active States 
and Sense State. 
 
*FristonBlanket is a subclass 
of ProbabilityRelation. 
*FristonBlanket is a subclass 
of Proposition.
 
 
Generalized 
Free Energy
Free 
Energy
Past Variational Free 
Energy plus future 
Expected Free Energy 
(each totaled over 
Policy).
 
*GeneralizedFreeEnergy is a 
subclass of 
ProbabilityRelation. 
*GeneralizedFreeEnergy is a 
subclass of Proposition.
 
 
Generative 
Model
Agents in 
the Niche
A formalism that 
describes the mapping 
between Hidden State, 
and Expectations of 
Action Prediction, 
Sensory outcome. 
Recognition Model 
Update Internal State 
parameter that 
correspond to External 
State (including Hidden 
State causes of 
environment states), 
Blanket State, and 
Internal State (meta-
modeling). In contrast, 
Generative Model take 
those same Internal 
State parameter 
Estimator and emit 
expected or plausible 
observations.
*GenerativeModel is a 
subclass of Process.
 
 
Generative 
Process
Agents in 
the Niche
Underlying @dynamical 
process in the Niche 
giving rise to Agent 
Observation and @agent 
Action Prediction.
Enactive ecological 
process using 
morphological computing
processes where the 
Niche Regime of 
*GenerativeProcess is a 
subclass of 
ProbabilityRelation. 
*GenerativeProcess is a

## Page 12

Attention 
@morphogenesis and 
generative model 
interact to create an 
embodied learning 
dynamic.
subclass of Proposition. 
*GenerativeProcess is a 
subclass of Process.
Hidden State
Markov 
Partitionin
g
Unobserved variable in 
Bayesian Inference, can 
reflect a Latent cause. 
 
*AbstractHiddenState is a 
subclass of 
ProbabilityRelation.
 
 
Hierarchical 
Model
Systems
A hierarchy of 
Estimators, which 
operate at different 
spatiotemporal 
timescales (so they track
features at different 
scales); all carrying out 
Predictive Processing 
 
*HierarchicalModel is a 
subclass of 
ProbabilityRelation. 
*HierarchicalModel is a 
subclass of Proposition. 
*HierarchicalModel is a 
subclass of Process.
 
 
Inference
Bayesian 
Statistics
Process of reaching a 
(local or global) 
conclusion within a 
Model, for example with 
Bayesian Inference.
The process of using a 
Sensory observation 
(observed variable, data)
along with a known set of
parameters to determine 
the state of an unknown, 
Latent cause 
(unobserved variable).
*AbstractInference is a 
subclass of Learning. 
(This looks wrong. Abstract 
classes are non-temporal, 
and Learning changes across
time.)
 
 
Information
Bayesian 
Statistics
Measured in bits, the 
reduction of Uncertainty 
on a Belief distribution of
some type. Usually 
Syntactic (Shannon) but 
also can be Semantic 
(e.g. Bayesian). 
 
 
 
*Information is internally
related to 
InformationMeasure.
Internal State
Markov 
Partitionin
g
States with #r65 
independent from 
External State, 
conditioned on Blanket 
State. 
 
*InternalState is a subset of 
PhysiologicProcess.
 
 
Learning
Bayesian 
Statistics
Broad sense: Process of 
an Agent engaged in 
Updates to Cognition 
(and possibly) Behavior.
Narrow sense: Process of 
Bayesian Inference 
where Generative Model 
parameters undergo 
Belief updating 
Learning is a subclass of 
intentional psychological 
process

## Page 13

Living system
Systems
Agent engaged in 
Autopoiesis 
 
 
 
*Attractor is internally 
related to 
SubjectiveAssessmentAt
tribute.
Logic
 
 
 
 
 
Logical operator is a 
subclass of predicate
Markov Blanket
Markov 
Partitionin
g
Markov Partitioning 
Model of System, 
reflecting Agent as 
delineated from the 
Niche via an Interface. 
The Markov Blanket 
Blanket State reflect the 
State(s) upon which 
Internal State and 
External State are 
conditionally 
independent. 
 
*ThermodynamicSystem is a 
subclass of *System
 
 
Markov 
Decision 
Process
Markov 
Partitionin
g
Bayesian Inference Model
where Agent Generative 
Model can implement 
Policy selection on 
Affordances reflected by 
Active States, while other
features of the 
Generative Process are 
outside the Control 
(states) of the Agent.
 
Deciding is a subclass of 
Selecting
 
 
Niche
Agents in 
the Niche
Ecology System 
constituting the 
Generative Process (as 
Partitioned from the 
Agent who instantiates a 
Generative Model). 
 
 
 
Attribute subsumes 
*Niche.
Non-
Equilibrium 
Steady State
Agents in 
the Niche
Technically, a Non-
Equilibrium Steady State 
requires a solution to the
Fokker Planck equation 
(i.e., density dynamics). 
A nonequilibrium steady-
state solution entails 
solenoidal (i.e., 
conservative or 
divergence free) 
Generally, a Non-
Equilibrium Steady State 
refers to a System with 
dynamics that are 
unchanging, or at 
Stationarity in some 
State. 
*NonEquilibriumSteadyState 
is a subset of Attribute.

## Page 14

dynamics that break 
detailed balance (and 
underwrite stochastic 
chaos).
Novelty 
Perceptio
n
The Internal State 
assumed by an Agent’s 
epistemic Affordance, 
when unable to 
immediately (e.g. locally)
resolve Uncertainty 
about the contingencies, 
i.e. the opportunity to 
resolve Uncertainty 
about ”what would 
happen if I did that?”
 
*Novelty is a subclass of 
SubjectiveAssessmentAttribu
te. 
SubjectiveAssessmentAttribu
te is a subclass of 
NormativeAttribute
 
 
Observation
Perceptio
n
The Belief updating of an
Internal State registered 
by a Sensory input, given
the weighting assigned 
to that class of input in 
comparison with 
weighting of the 
competing Priors. (This is
a narrow sense of 
“observation,” where the
Agent is “looking for this 
kind of input.”)
Any Sensory input, either
discrete-valued or 
continuous-valued. (This 
is a broad sense.)
 
 
*Observation is 
internally related to 
CognitiveAgent. 
Perception
Perceptio
n
Posterior State Inference 
after each new 
Observation.
 
Perception is a subclass of 
psychological process
 
 
Policy
Action
Sequence of Actions, 
reflected by series of 
Active States as 
implemented in Policy 
selection which is Action 
Prediction or Action and 
Planning as Divergence 
Minimization 
 
Policy is a subclass of 
Proposition.
 
 
Posterior
Bayesian 
Statistics
The Update to the Prior 
after Observation has 
occurred 
In Bayes”' theorem, the 
Posterior is equal to the 
product of the Likelihood 
and Prior divided by the 
model evidence.
*Posterior is a subclass of 
Proposition.
 
 
Pragmatic 
Free 
Pragmatic Value is the 
Pragmatic value 
*PragmaticValue is a 
 
*PragmaticValue is a

## Page 15

Value
Energy
benefit to an organism of
a given Policy or Action, 
measured in terms of 
probability of a Policy 
leading to Expectation of 
Random variable values 
that are aligned with the 
Preference of the Agent 
describes the extent to 
which a given action is 
aligned with rewarding 
preferences over sensory
outcomes.
subclass of StateOfMind.
relatedInternalConcept 
to Selecting.
Principle
 
Principle 
 
*Principle is an instance of 
Proposition.
 
*Principle is internally 
related to Reasoning. 
Reasoning is an instance
of Proposit 
IntentionalPsychological
Process
Propositional 
attitude
 
 
 
attitudeForFormula is an 
instance of ternary relation. 
The number 1 argument 
of attitudeForFormula is 
an instance of 
EmotionalState.
The number 2 argument 
of attitudeForFormula is 
an instance of agent.
The number 3 argument 
of attitudeForFormula is 
an instance of Formula. 
 
Recognition 
Model
Agents in 
the Niche
Recognition Model is the 
kind of Model that affords
Variational Inference, 
which lets us calculate or
approximate a 
probability distribution. 
Recognition Model is a 
synonym for Variational 
Model.
In Dayan and Abbot 
(2001), the probability of 
a Hidden State (causes) 
given Sensory Data 
(effects) under some 
parameter. 
Recognition Model is a 
subclass of Modeling.
Modeling is a subclass of 
intentional process
 
*Recognition is 
internally related to 
Realization. 
Recognition Model is 
internally related to 
Variational.
Representation
Agents in 
the Niche
A structural 
correspondence between
some Random variable 
inside a System and 
some Random variable 
outside the System 
(isomorphism being the 
strongest kind of 
 
represents is a subrelation of
refers.
refers is a relation from an
Entity to an Entity. 
* If a process is an 
instance of expressing 
and an agent is an 
agent of the process,
* then there exists an 
attribute such that the 
attribute is an instance 
of state of mind and the

## Page 16

correspondence), such 
that the System engages
in Inference carried out 
by the System maintains 
the correspondence
attribute is an attribute 
of the agent and the 
process representsthe 
attribute
represents 
 
SUMO relation 
(represents ?THING ?
ENTITY) means that ?
THING in some way 
indicates, expresses, 
connotes, pictures, 
describes, etc. ?ENTITY. 
The Predicates 
containsInformation and 
realization are 
subrelations of 
represents.
 
represents is an instance of 
binary predicate.
represents is a subrelation of
refers. 
The number 1 argument 
of refers is an instance of 
Entity.
The number 2 argument 
of refers is an instance of 
Entity. 
 
Sense State
Markov 
Partitionin
g
In the Friston Blanket 
formalism, the Blanket 
State are the Sense State
(incoming Sensory input) 
and Active States 
(outgoing influence of 
Policy selection )
 
*SenseState is a subset of 
PhysiologicProcess
 
 
State
Bayesian 
Statistics
is the statistical, 
computational, or 
mathematical value for a 
parameter within the 
State space of a Model. 
 
Attribute is a subclass of 
abstract.
 
“State” is a near 
synonym of Attribute. 
State space
Bayesian 
Statistics
Set of 
variables/parameters 
that describe a System.
A state space is the set 
of all possible 
configurations of a 
system
*StateSpace is a subset of 
Attribute
 
 
System
Systems
Set of relations described
by State space of a 
Model. 
Differentiable and 
Integratable in terms of 
Variables and functions.
*System is a subclass of 
Agent
 
 
ACKNOWLEDGMENT
To the late Prof. Gordon Pask, who taught me how to think about thinking.

## Page 17

REFERENCES
Douglass, David S. et al. (2024). Active Inference Topics Aligned to SUMO. DOI 10.5281/zenodo.11459323
Douglass, David S. et al. (2024). Aligning Spatial Web Terminology to SUMO. https://zenodo.org/doi/10.5281/zenodo.11055372
Pask, Gordon and Bernard C.E. Scott (1973). CASTE: A System for Exhibiting Learning Strategies and Regulating Uncertainties. Int. J. Man-Machine Studies (1973) 5, 17-52. 
https://www.sciencedirect.com/science/article/abs/pii/S0020737373800082?via%3Dihub
Niles, I., & Pease, A., (2001). Toward a Standard Upper Ontology, in Proceedings of the 2nd International Conference on Formal Ontology in Information Systems (FOIS-2001), Chris Welty and 
Barry Smith, eds, pp 2-9. https://dl.acm.org/doi/10.1145/505168.505170
Pease, A., (2011). Ontology: A Practical Guide. Articulate Software Press, Angwin, CA. ISBN 978-1-889455-10-5
Pease, Adam (2024). Suggested Upper Merged Ontology (SUMO). https://ontologyportal.org/


---
*Extraction method: pymupdf*