# Full Text: Active Inference & Behavior Engineering for Teams

> Extracted from `2020_BehaviorEngineering.pdf`

---

## Page 1

● Communication is fundamental in natural and designed teams; effective communication 
systems are needed to facilitate the organization and function of complex multi-agent 
systems. 
● Under the Free Energy Principle (FEP), previous work has synthesized Active Inference 
(ActInf) with domains such as Narratives [1], Ontologies [2] and extended cognition in 
multiscale biological systems [3,4]. 
● Using the ActInf framework, here we explore various kinds of Communication in located 
teams and all-online Remote Teams (RT).  
● Online work and RT are promising systems for theoretical study and direct applications of 
the ActInf framework because all states and updates in digital systems are observable. 
● Here we bring the FEP-ActInf-Narrative nexus together with the applied approach of 
Systems Engineering (SE), to begin the work of formalizing the processes of RT formation 
and lifecycle management.

## Page 2

● How could we consider the coherence, narrative, and identity function of communication 
at the individual scale, as well as the scale of Teams & organizations? 
● What kinds of methods for Teams (analytics, user interfaces, etc) could be deployed to 
address basic and applied questions of interest?  
● How could we apply multi-scale Active Inference frameworks to Systems approaches such 
as Ontologies, Narratives, Formal documents, and Tools (ONFT)? 
● How could we address concepts and models for epistemic values within the context of 
ONFT for Remote Teams (epistemic foraging)? 
● How could Teams ensure narrative reliability? How can the epistemic and goal-oriented 
ends of foraging be jointly optimized by individuals and teams?  
● What we could take from concepts and works on niche construction to aid development 
of ONFT approaches for modern, global Teams? 
● Could we define approaches for personal behavior engineering by using ONFT in Teams 
communication?  
1. Teams are about function & communication 
2. Modern Teams are Remote Teams 
3. Systems Engineering provides frameworks for life cycle management of complex systems 
4. Active Inference in Teams 
Work is performed by teams, human and non-human (e.g. ants) [5–7]. The concept of Division of 
Labor describes how system subunits interact with each other and perform work [7,8] in Complex 
Adaptive Systems regulated by agent-agent and agent-environment feedback systems [7,9–11] 
(also see Task Allocation [12,13], Heterarchy [14–16], and ant semiotics [6,17,18]). In the context 
of remote and located Teams, heterarchical subsets of members and stakeholders allocate tasks 
based on practices (norms) and Roles (identities or assignments). Team members not only 
perform work, but they also send signals, exchange results, and contribute to shared and 
documented models in extended cognitive tools [19,20]. Tools can help both long-lasting and 
rapidly-assembling Teams deal with challenges which can’t be solved or resolved by any single 
person. For example, good Team documentation software enables efficient usage of distributed 
expertise & transdisciplinary cognition, via allowing the affordance of interacting with the wisdom 
of previous teammates [21,22].  
For humans, narratives are aspects of individual and shared generative models of the 
world [23]. For teams, multi-scale narratives emerge as individuals build a generative model of

## Page 3

their team. We highlight the phenomenological experience of an individual worker as they 
investigate high level narratives (Why does this company exist? What problems is it solving or 
impacting in the world?) as well as team-level narratives (Why does this team exist? Why is our 
part of the project important to the whole company?), Narratives become memetic when they can 
be shared and underwood in common, this process of communication leads to alignment based 
upon shared values, mission, etc. 
There is a need to define Team communication in a more formal fashion, ideally drawing on 
insights from transdisciplinary theoretical (e.g. Complexity Science) and applied (e.g. Systems 
Engineering, Systems Innovation) approaches [24,25]. Modern Teams are composed of sets of 
human, collective, or non-human agents [26–29], often with high turnover rate (see Definitions) 
[21,30,31]. Today’s Remote Teams (RT) are physically distributed, and increasingly use the 
internet to coordinate action and informationally connect team members [24,32,33]. When 
working with Instantaneous Remote Teams (IRTs), one also needs to consider the timing at a 
very fine scale – in IRTs, each team member could participate in different Teams in different 
capacities during one working day or even a single hour [31]. 
High Reliability Organizations (HROs) are organizations that contend with volatile 
environments in which many interactions can be considered non-routine. HROs are increasingly 
reliant on small, physically distributed, and sometimes temporary or rapidly assembled teams as 
a means of solving novel, complex problems [34–37]. Examples of these Teams include “tiger 
teams” in the oil and gas industry [34], and “swift market analysis response teams” (SMART) in 
the auto industry. Key rituals, protocols, and strategies for these IRTs have been incorporated 
into SCRUM and Agile Development frameworks for rapid development of software as well [37]. 
The rapidly assembled, or sometimes “instantaneous” remote team, is an emergent solution to a 
set of emergent problems. Human knowledge has expanded exponentially, consequently, fields 
of expertise began to divide into specializations as a basis to reduce time-to-application and 
learning requirements [38]. This subspecialization has achieved its goals at a cost: generally, no 
single individual and often no single team holds all of the knowledge and skills necessary to solve 
the novel problems emerging from the complex threat surfaces with which HROs contend, and 
as a result, reconfiguration is becoming a more normalized response [25,39,40]. This solution 
isn’t unique to industry: National Navies are organizations which contend with complex threat 
surfaces in littoral environments, with the additional constraint that equipment repertoires are the 
product of decade-long investment cycles [25,41,42]. Consequently many National Navies have 
converged on the same outlook: that no single team or equipment configuration is adequate for 
the future of expeditionary warfare and indeed remote work [41–43].  
Where emergent teams of any type are created in response to novel, complex problems, 
they cannot rely on effective informational compression via inflexible protocols, as the team

## Page 4

situation and even composition may be a moving target. Rapidly-forming teams sometimes are 
precluded from relying on compression via long-term bonding as would be found in traditional 
high performance teams. As such, it should be unsurprising that high performance, emergent 
teams responding to novel, complex problems generally rely on shared organizational culture, 
mission, and narrative [25,31,37,44–49]. In the context of both internet communities & in-person 
protests, memes can serve as rallying points as well as symbols that communicate mission-critical 
narrative information to team members [25,31,50]. Organizational culture in remote and located 
teams can be defined as the shared beliefs and values of an organization, as well as its collective 
processes, cognitive and physical [51–53]. Mission has three primary connotations: military 
[54,55], religious [56,57], and corporate [58,59]. In all three usages of “mission”, mission-relevant 
narratives and symbolic (or even esoteric) communication are used for the purpose of 
compressed goal-setting.  
In the context of team communication, narratives are dynamic, and in constant adjustment 
[38,60,61]. Narratives become recognizable through shared or attuned semiotics, iconology, and 
totemization [47,62–66]. Narratives can be created, perturbed, and managed [63] through the 
production of physical artifacts [67] as well as through ritual [68–70]. Narratives are a form of 
memetic compression, for example Linn’s reduction of three centuries of American military 
philosophy into three “camps”; Guardians, Heroes, and Managers [71]. This concept of “narrative 
as dynamical analogy” is about finding the stable mappings within complex systems that allow for 
effective action, as is sometimes used in physics [72], computer science [73], and in memetics 
itself [74,75]. 
Rapidly formed teams and IRTs come together with clear limitations, the most important of 
which is the social cohesion and trust necessary for organizational sensemaking. Effective 
formation of small teams leads to optimal utilization of collective intelligence, and generally 
positive performance [37,76–81]. Conversely, failure to develop mutual trust and social cohesion 
can hinder performance [80,82–84]. When opportunities for a team (startup, governmental, 
research, or otherwise) are dynamic and require rapid reorientation, failures of team formation 
can be lethal [37,51,84,85]. Teams have both implicit and explicit organizational structures & 
networks of communication. These defined or undefined team structures (representations of 
networks of roles, positions, signals) have direct implications for the efficacy of communication 
and production of Team artifacts (physical, software, narrative, memes). Functional small teams 
can be argued to belong (exclusively or non exclusively) to at least one of three classifications 
characterized by the means by which members reduce uncertainty about the signals and actions 
of other members, presented here:  
● ONTOLOGICAL ALIGNMENT: The first kind of group is composed of organizations which 
depend on very strict, clearly defined, compressed ontologies paired with strict processes 
that limit the potential for signal-error, creating high expectations of trust between

## Page 5

individuals who do not necessarily know each other or even expect to interact again, such 
as operating rooms, or air traffic control [34,51,86–88].  
● INTIMATE TRUST ALIGNMENT: The second kind of group is found within organizations 
which depend on team bonding and practice over very long periods of time in order to 
create high-trust and “short-hand” communication that is very highly compressed even if 
ambiguous or indiscernible to external actors – this group includes organizations which 
create a “collective mind” during operations such as special operations units, fire 
departments, sports teams, aircraft carrier flight decks, and non-human cooperative 
hunting groups such as wolves [44,51,87,89,90].  
● NARRATIVE ALIGNMENT: The third is composed of organizations which are aligned on 
organizational culture, narrative or mission [31,37,46,91,92]. These groups can be 
anonymous, and dynamic in composition or focus. 
Here, by defining Teams in terms of their communicative structure, we include many informal 
groupings (internet chat rooms, crowds, protests, spontaneous public meetings) not classically 
considered as Teams. Our Team definition here is oriented towards capturing the diversity of 
communicating systems, rather than just the explicit organizational structures. In modern 
contexts, teams assemble and disassemble over short timescales, and are often composed of 
not just of humans but also non-human facilitation agents [26–29,93]. What is not a team, under 
this conception? The short answer to this question, which will be explored later in the context of 
Team Markov Blankets, is that non-communicating entities, or entities that are not part of the 
same informational niche, are not part of the same team. Non-communicating entities may still 
have alignment of values, mission, or even behavior – but they are not on the same team because 
they are in non-overlapping informational niches. All of these examples point to the need for a 
formalized system for today’s RT that can meaningfully cope with all of these strategic and tactical 
challenges. 
To take a field from theoretical speculation to applied utility, we need a set of tools for defining 
and interacting with a System of Interest (SoI). In this case we are interested in behavior 
engineering of team narratives communication structures. Engineering is always about changing 
something in the world, and so behavior engineering in teams is no exception. Engineering can 
benefit from the Complexity Science perspective, but a conceptual approach alone is incomplete 
for the designing and implementing of real systems. To quote from the definitive Systems 
Engineering (SE) Book of Knowledge: “Systems Engineering (SE) is an interdisciplinary approach 
and means to enable the realization of successful systems. It focuses on holistically and 
concurrently 
understanding 
stakeholder 
needs; 
exploring 
opportunities; 
documenting

## Page 6

requirements; and synthesizing, verifying, validating, and evolving solutions while considering the 
complete problem, from system concept exploration through system disposal” [94].  
SE frameworks define usage of Division of Labor for life cycle management based on 
different functional Roles for each stage. For each task to be performed, Practices (e.g. 
architecture, development, testing) are supported through technologies relevant to each Role. SE 
defines “Practices” as the combination of discipline, work, products, tools and activities [95,96]. 
To provide actionable solutions to pressing needs, Systems Engineering defines the “functions” 
and objects of attention during work on SoI life cycle [95,96]. In SE, functions are also a key unit 
of analysis. These functions can be carried out by multiple humans, or one human may have 
multiple roles/functions. Thus the design imperative, within a Division of Labor context, is to 
configure the roles in order to produce functional outcomes. This refocuses the discussion away 
from spurious communication, and toward task-oriented or performance-oriented outcomes. The 
pursuit in SE of expected team outcomes is akin to the cybernetic idea that complex self-
regulating systems must be goal-seeking in order to survive and thrive [97–99].  
We draw on the OMG Essence framework [100] to explore the use of Alphas (Abstract-
Level Progress Health Attribute), which are uncertainty-reducing sets of States and Checklists for 
that track changes in the performance of collective work. Teams, at any given moment, are 
focused on a single Alpha that rises to the level of group attentional awareness [101], akin to the 
emergence of high-level salience in hierarchical systems [4,102]. The Essence framework 
identifies seven Alphas as objects of attention in every software engineering project: 
Stakeholders, Opportunity, Requirements, Software System, Team, Way of Working, and Work 
[100]. These seven categories also apply well to RTs. In the course of the project the status of 
the team undergoes small and large changes, passing through states as work is performed. 
These states of teams and products are observable, in contrast to Alphas, the states of which we 
can only judge "by instruments"–by the state of artifacts. 
It was proposed in Systems Engineering Essence framework [95,96] to expand 
applications of Alphas from software projects only to hardware and sociotechnical projects by 
changing Requirements and Software System to System Definition and System Realization. To 
capture some of the useful ideas from SE, we summarized several recent summary documents 
of Systems Engineering (Table 1 & Table 2). Despite the fact that SE approaches well established 
and have been used widely in the last decades it is still the general opinion that SE needs to 
interface with people outside the scope of a system, even though there is no way to directly 
engineer their behavior. One possible solution to this challenge of integrating internal and external 
SoI dynamics would be to set patterns and rules for internal and external communication [103]. 
We now turn to the enactive framework of Active Inference to provide inspiration for the design of 
communication patterns for RT that would facilitate modern teamwork.

## Page 7

Active Inference (ActInf, see Definitions) is a formal framework that frames goal-seeking behavior 
as an actor-centric dynamic feedback between internal and external states, mediated by sense 
and action [104–106]. ActInf is a process theory (as opposed to a state or variance theory 
[107,108]) based upon the Free Energy Principle [102,109] (Figure 1). In ActInf, generative 
models about the world (as opposed to descriptive, reactive, or analytical models) support 
ecologically-relevant functions of real systems [3,97,110] – for example a person trying to catch 
a ball with move towards where they predict the ball will intersect with their trajectory, and motor 
saccades of the eye during reading are related to real-time predictions about which visual 
information will be most informative [111,112]. ActInf captures informational and statistical 
aspects of these generative models and how they are updated and communicated by multiscale 
far-from-equilibrium systems [104–106,113,114]. ActInf thus presents as a promising approach 
to the quantitative study of complex system behavior [3,97,110]. In this article our focus is on 
situating team communication as a case of Active Inference, and exploring various avenues 
where ActInf approaches could be useful for modern teams.

## Page 8

Here we briefly review several recent developments in the ActInf literature that are relevant 
for our use case of RT. The topics of communication, narrative, and culture have recently been 
contextualized within the context of ActInf and the FEP [116–119]. Communicating systems such 
as the brain [120,121] and improvising dyads [122] can be formally cast within the ActInf 
framework, making these varied systems amenable to powerful physics-based analyses. For 
humans, the study of semantic interpretation of text is known as hermeneutics, which lies at the 
base of many forms of communication. ActInf captures how multiple interacting agents perform 
improvisational hermeneutics at the behavioral timescale (via e.g. micro-scale turn taking [122]), 
scaffolded within cultural niches that play out at much longer timescales [123]. The expected 
status of communication within human teams is cooperative, facilitating the emergence of 
effective work on large and complicated projects [115].  
In the case of goal-oriented team work, ActInf explores how agents communicate with 
each other in order to reduce each other’s uncertainty about internal (mental) and external (world) 
states. In order to coordinate at higher scales, agents must be connected through communication 
channels (shared epistemic niche) as well as have the Bayesian prior belief that attunement or 
alignment is a desirable outcome (desirable since it would reduce uncertainty about achieving 
preferred future sensory states) [115]. Over evolutionary time, the priors that communication 
among similar agents is cooperative becomes entrenched through selection (assuming that 
coordinating agents have higher fitness) [4]. These evolutionary and developmental expectations 
about social interactions are enacted and shaped through real-time experience – giving a formal 
sense to the classic phrase “through others we become ourselves” [124,125]. We can adapt this 
phrase here to consider how teams form and perform, e.g. “through communication with others

## Page 9

we become a team”, or “though reducing our uncertainty about the future we achieve our shared 
goals”.   
The rest of this paper is dedicated to exploring the features and implications ONFT for 
teams, using SE and the ActInf framework. We focus on the multiple levels of communication that 
teamwork entails (within and between teams), and some of the special aspects of modern Remote 
Teams (for example rapidly changing composition and augmented or non-human teammates). 
We explore the possibility of creating protocols for RT communication to succeed in the 
development of SoIs, based upon optimized message passing systems inspired by ActInf. This 
perspective for computational communication of RT extends naturally from recent work on ActInf 
in enactive and encultured communication. 
Here we highlight the Remote Team (RT) as a tractable "model system" for studying the 
processes of Communication, Narrative co-construction, collective intelligence, organizational 
sensemaking, and organizational management online. The design of successful RTs, now more 
than ever, is essential for the health and productivity of modern society circa 2020. We use the 
multiscale action-oriented framework of ActInf to consider the communicative, psychological, and 
techno-social dynamics of RTs [4,97,112]. We consider how recent developments in online 
organization, gamification, and platform accessibility make formal systems for RT & 
Instantaneous Remote Teams (IRTs) a relevant technology for research and implementation [25]. 
Overall we aim towards Ontologies, Narratives, Formal documents, and Tools (ONFT) for RT 
within the ActInf framework.  
First, we can recast the generalized AcfInf setting of Figure 2 into the specific case of two 
(or more) interacting team mates within a shared team informational niche (Figure 3). We use the 
concept of a Markov Blanket (MB, see Definitions) and communicating systems to define a team 
as the set of human and non-human agents that share a specific informational niche (Figure 4). 
In ActInf, the MB reflects the separation between internal and external system states, pierced by 
active and sensory states [126–128]. In the context of team communication, the MB is enacted 
by the informational boundaries of the team, though there may also be permanent or transient 
internal subdivisions [129], especially in large organizations with reconfiguring subteams. 
Communication among team members in RT can take various forms (Figure 5), including audio-
visual relay (video chat), text messages (chat), file sharing, and other forms of information transfer 
(sensors, biofeedback, geospatial information).

## Page 10

*[Page 10 appears to be blank or image-only]*

## Page 11

*[Page 11 appears to be blank or image-only]*

## Page 12

We consider narratives as tools that have functional roles in Teams (Figure 6). At the 
same time narratives can be generative models (enacted or latent generative dynamics). Our 
focus is on team behavior, in that we want to attune communication in the interest of achieving 
certain results. This focus on quality production of artifacts will be behaviorally accomplished in 
real teams through the design of effective communication and regimes of attention. This is 
consistent with recent developments in ActInf which frame cultures (of organizations, teams) as 
“cultural scaffolds” and “regimes of expectations” [97] that through communication are able to 
achieve higher-order goals [105]. 
 
Remote Teams (RT) are especially tractable for formal analysis of any kind, because most 
state transitions in the team are observable. In located teams, it can be challenging to capture the 
nuance of important communication techniques such as space use or body language. Conversely 
in an RT, while body language and other qualitative ostensive cues may still be critical, an 
observer can be sure that they are at least capturing all of the signals being exchanged (unlike, 
for example, a video camera in a conference room which may be able to capture where each 
person is in the room, but not what each person sees). Continuing with this mapping between RT 
and other far-from-equilibrium message passing systems graphs that perform Active Inference, 
we can consider all agents (human or non-human) as nodes that are connected via 
communicative edges. The structure of this graph is the realized communication system of the 
team, and thus the boundaries of the work-performing aspects of the team. Nodes that are 
informationally connected may be formally related (e.g. via an Org chart) or they may be 
organizationally unlinked.  
Different kinds of communicative edges may reflect different types of relationships such 
as informant or close interpersonal linkage (friendship or “buddyship”, reflecting a highly 
synchronized shared generative model). In the RT, because all communications are via online 
transfer, this exocortex is a key Enabling System [130,131]. This means that we can define the 
Enabling System in terms of all communication events (in Online space), and for each event 
provide defined roles & protocols. Narratives for communication are essential for all sorts of team 
relationships. Narratives can be social functions that create a cognitive niche, thus reducing 
collective/individual uncertainty [115,132]. Within the context of a narrative that sets the team goal 
and function, there is a process of exocortex-driven Division of Labor. Functional ontologies are 
relevant for the role-assignment stage, whereas the System-level ontology steps in to help 
workers make sense of what they should do.  
We see Active Inference as something like a “two stroke engine” for Remote Teams (Act 
–> Infer –> Act –> …), accomplished through the communicative structure or “Syntax” of the RT 
(also see OODA loops [133,134]). In all cases, everything is based upon, or supported by tools. 
This means that work is performed through observable sequences of Events (taking place at a 
specific time with specific syntax/grammar) which result in meaningful progressions of events 
(narrative semantics). For an event to exist, there must be a measurable change in a system state

## Page 13

or Alpha [135]. From the perspective of the team members, communication about narratives is of 
the utmost importance, as narratives set the stage for interpretation of subsequent signals. 
Narratives are strong enough to serve as nucleating or rallying points for located protests as well 
as all-online IRTs, underlining the need to understand how memes and narratives interact in 
modern informational ecosystems [25]. 
 
Here we use the framework of ONFT to highlight specific areas where ActInf could be 
applied:  
 
1. Ontologies: 
a. Active Inference could inspire action-oriented ontologies for Remote Teams, 
describing team composition, communication systems, work performance, 
informational channels, hardware/software, and more. This leads to the idea of 
interoperable RT from different organizations (e.g. using standards for metadata 
that allow for data transformation and Business, Operations, Legal, Technical, and 
Social inter-team communications and situational awareness such as those 
proposed through work on Coalition Battle Management Language [136–138])).  
i. 
Required ontological information for team communication could include 
(Date, Time, Sender, Role, Alpha). Optional information could include 
(Seals, Symbols, Context, and Signposts for regimes of attention). 
b. An ActInf-based ontology for Narratives would allow the design or control of 
narratives in RT. This might be facilitated by tools like Sentiment analysis, 
visualization techniques, and machine learning of social media data. Other 
researchers have sketched out common cases where narratives for online teams 
already exist, how could a formal structure make this more manageable? 
c. In terms of Team membership and informational ingresses/emissions, ontologies 
for multi-agent systems and Markov Blankets might allow for the design of internal 
and external representations of work performance [2]. 
 
2. Narrative: 
a. Narrative alignment is dynamic and grows between members of teams through 
peer or “horizontal” bonding [84]. Organizations which consist of many teams may 
experience narrative alignment via both horizontal and vertical bonding, that is, 
bonding with team-mates and members of other teams as well as bonding with 
supervisors [84]. The highest level of narrative alignment might be best described 
with the military term “esprit de corps”, where mutual sense of mission, trust, ideals, 
culture, and shared threats allow alignment to transcend self-interest, specific unit 
membership, and limits on intimate relationships [84,139–142]. Narrative 
alignment associated with “esprit de corps” creates behavioral ideals, objectives,

## Page 14

heroic tradition, and culture or “regimes of expectations” for individuals to align and 
conform 
with 
in 
order 
to 
cope 
with 
high 
levels 
of 
uncertainty 
[1,4,71,83,97,142,143]. ActInf frameworks for RT could promote a digital “esprit de 
corps” that is observable and also tractable to interface with.  
b. The value of communication patterns in the RT could be quantified in terms of 
value for the team narrative, as proxied by novel evidence for (updated distribution 
of) shared generative models. This is similar to how backpropagation training of 
neurons in an artificial neural network updates parameters based upon contribution 
to error, or how Numer.ai rewards machine learning models proportionally to how 
they contribute to the success of an automated trading bot [144].  
c. At different levels, we can associate different functions for different generative 
narratives of interest. We need to be able to name, trace, and document the states 
of narratives (as well as capture pluralistic interpretations of multi-person 
narratives).  
d. A focus on function and role performance within a narrative context could improve 
the performance of work and the experience of team members. This is because 
narratives are functions that provide Identity and Meaning across multiple scales. 
From a SE perspective, Narrative is just another SoI that we can reduce our 
uncertainty about, towards the end of system design and cybernetic control. Just 
as with other Complex control questions, we are able to design/control at the 
Systems level by making the right abstraction for coarse-graining (here, Markov 
Blankets that allow us to ignore hidden internal states).  
 
3. Formal Documents:  
a. ActInf could inform the design of documents that relate multiscale event 
frameworks – Each event has prerequisites, inputs/outputs (functionalism), 
consequences & outcomes, Roles, Problems in focus, expectations, and 
predictions. Formal documents capture which engineering metadata needs to be 
present (e.g. reference data format) in order to perform life cycle analysis on SoI.  
b. Formal Documents for the Work Day and Week could improve the experience of 
workers.  
i. 
Morning documents: providing narrative alignment and informational 
update for the day. 
ii. 
End-of-day documents: providing closure to the day, filling out information 
about progress. 
iii. 
Monday documents: providing narrative Alignment for the week (mission, 
culture, identity, collective sense-making, where are we in the bigger 
niche?)  
iv. 
Friday documents: providing closure for the work week.

## Page 15

4. Tools: 
a. Tools are required in all of the above domains so that professional, innovative, 
effective, inclusive Remote Teams can implement effective ActInf frameworks of 
any kinds. Current common (and often free) tools include chat, file-sharing, 
voice/video, CRM, Single Source of Truth software, etc. Such tools will be helpful 
for ActInf-based teams, and also new kinds of tools may be required.  
b. Given the total observability of RT, toolkits such as SPM [145,146] and multiscale 
analytics could help attune RT communication towards desired products. Human-
in-the-loop machine learning systems based upon ActInf could allow for actions 
and perceptions to be designed and controlled in real-time at a fine scale 
[104,147].   
c. For RT communication across platforms, it would be helpful to design common 
database formats that link protocols, for example using an API connector like 
Matterbridge 
[148]. 
This 
would 
allow 
a 
nuanced 
tradeoffs 
between 
centralized/private/decentralized backends that used custom metadata, and user-
facing platforms with customizable UI/UX and dynamic data updating. This kind of 
“total comms” understanding, and ability to design effectively within the space of 
possible RT, would reduce platform fragmentation and increase worker 
effectiveness.  
d. Inspiration from nature (biomimicry) could provide new tools and perspectives on 
how different work functions could be performed by different cognitive niches 
[6,149]. 
e. Computer-assisted design (CAD) Tools for communication charts would allow the 
formalization of “Markov communicative blankets”. This could facilitate the 
formation of collective cognitive entities that can then be understood, compressed, 
templated, optimized, and reconsidered from multiple perspectives [150].   
f. 
Tools for regimes of synchronous & asynchronous attention would allow for the 
optimal design of ostensive cues and salient epistemic signals – “events only 
happen when the listener is paying attention”.  
The future of the Free Energy Principle and Active Inference is bright but uncertain. Through our 
cybernetic communication and actions in the now, we reduce our uncertainty about the hereafter.

## Page 16

System Fundamentals 
System 
Behavior 
Complexity 
Emergency 
Representing Systems with Models Model 
Model-based Systems Engineering 
Modeling Language 
Engineered System Context 
Product Systems 
Service Systems 
Enterprise Systems 
System of Systems 
Cyber-Physical Systems 
Systems Engineering Standards 
Modeling Standard 
Related Standard 
Generic Life Cycle Stages 
Concept Stage

## Page 17

Development Stage 
Production Stage 
Utilization Stage 
Support Stage 
Retirement Stage 
Systems Engineering Management Planning 
Assessment and Control 
Decision Management 
Risk Management 
Configuration Management 
Information Management 
Measurement 
Quality Management

## Page 18

Alpha 
State 
Description 
Stakeholders 
The people, 
groups, or 
organizations 
who affect or are 
affected by a 
system. 
Recognized 
Stakeholders have been identified. 
Represented 
The mechanisms for involving the stakeholders are agreed 
and the stakeholder representatives have been appointed. 
Involved 
The stakeholder representatives are actively involved in the 
work and fulfilling their responsibilities 
In Agreement The stakeholder representatives are in agreement. 
Satisfied for 
Deployment 
The minimal expectations of the stakeholder representatives 
have been achieved. 
Satisfied in 
Use 
The system has met or exceeds the minimal stakeholder 
expectations. 
Opportunity 
The set of 
circumstances 
that makes it 
appropriate to 
develop or 
change a 
software system 
Identified 
A commercial, social, or business opportunity has been 
identified that could be addressed by a software-based 
solution. 
Solution 
Needed 
The need for a software-based solution has been confirmed. 
Value 
Established 
The value of a successful solution has been established.

## Page 19

Viable 
It is agreed that a solution can be produced quickly and 
cheaply enough to successfully address the opportunity. 
Addressed 
A solution has been produced that demonstrably addresses 
the opportunity. 
Benefit 
Accrued 
The operational use or sale of the solution is creating 
tangible benefits. 
System definition 
A set of core 
technical 
activities of 
systems 
engineering, 
including the 
activities that are 
completed 
primarily in the 
front-end portion 
of the system 
design. 
Conceived 
It is clear how the system will be defined. 
Consistent 
Consistent System definition has been created. 
Coherent 
The requirements provide a consistent description of the 
essential characteristics of the new system. 
Used for 
Production 
System definition is used for system production. 
Used for 
Verification 
System definition is used for testing. 
Used for 
Operation 
System definition is used by stakeholders for operation. 
Used for 
Disposal 
System definition is used for system disposal. 
System 
Realisation 
The activities 
required to build, 
Raw materials Raw materials for system realization are available and allow 
manufacturing of the parts with required properties. 
Parts 
Parts have been produced and are ready for integration.

## Page 20

integrate 
elements, and 
ensure that a 
system both 
meets the needs 
of stakeholders 
and aligns with 
the requirements. 
Demonstrable The system has been assembled from the parts and is ready 
for testing. 
Ready 
The system (as a whole) has been accepted for deployment 
in a live environment 
Operational 
The system is in use in an operational environment. 
Retired 
The system is no longer supported and disposed and/or 
recycled. 
Team 
A group of people 
actively engaged 
in the 
development, 
maintenance, 
delivery, or 
support of a 
specific software 
system. 
Seeded 
The team’s mission is clear and the know-how needed to 
grow the team is in place. 
Formed 
The team has been populated with enough committed 
people to start to pursue the team mission. 
Collaborating The team members are working together as one unit. 
Performing 
The team is working effectively and efficiently. 
Adjourned 
The team is no longer accountable for carrying out its 
mission. 
Way of Working 
The tailored set of 
practices and 
tools used by a 
team to guide 
and support their 
work. 
Principles 
Established 
The principles, and constraints, that shape the way-of-
working are established. 
Foundation 
Established 
The key practices, and tools, that form the foundation of the 
way of working are selected and ready for use. 
In Use 
Some members of the team are using, and adapting, the 
way-of-working.

## Page 21

In Place 
All team members are using the way of working to 
accomplish their work. 
Working well 
The team's way of working is working well for the team. 
Retired 
The way of working is no longer in use by the team. 
Work 
Activity involving 
mental or 
physical effort 
done in order to 
achieve a result. 
Initiated 
The work has been requested. 
Prepared 
All pre-conditions for starting the work have been met. 
Started 
The work is proceeding. 
Under Control The work is going well, risks are under control, and 
productivity levels are sufficient to achieve a satisfactory 
result. 
Concluded 
The work to produce the results has been concluded

## Page 22

● RT – Remote Team 
● IRT – Instantaneous Remote Team 
● MB – Markov Blanket 
● SoI – System of Interest 
● ActInf – Active Inference 
● FEP – Free Energy Principle  
● ONFT – Ontologies, Narratives, Formal documents, and Tools  
● Definition: “Active Inference” (ActInf)  
○ Active inference (ActInf) is an action-oriented process theory that is related to the 
formal multiscale framework of Free Energy Principle (FEP) [3,109]. ActInf posits 
that action-perception cycles link external and internal systems, through sensory 
and active states that bidirectionally constitute a system-specific boundary known 
as a Markov Blanket (MB). ActInf is related to areas such as Cybernetics [152], 
Niche construction [110], Information foraging [111], linguistics [105], Variational 
Bayesian machine learning techniques [104].  
● Definition: “Engineered System”  
○ An open, concrete system of technical or socio-technical elements which is the 
focus of a SE life cycle. Its characteristics include being created by and for people, 
having a purpose and satisfying key stakeholders’ value propositions when 
considered as part of a broader system context [94].  
○ An engineered system is a system designed or adapted to interact with an 
anticipated operational environment to achieve one or more intended purposes 
while complying with applicable constraints [153]. 
● Definition: “System life cycle” 
○ The evolution of a system, product, service, project or other human-made entity 
from conception through retirement [95,154]. 
● Definition: “System of Interest” (SoI)  
○ The system whose life cycle is under consideration [154]. 
● Definition: “Team”  
○ A set of communicating nodes, where nodes represent actors (people, augmented 
people, computers). Teams with coherence (of communication, narrative, or 
function) tend to be involved in a shared work. The performance of this functional 
work is in feedback with Team informational connectivity, as well as the extent of 
attunement of shared beliefs, policies, goals, values, and worldview among 
stakeholders.  
○ Team composition and mission are all subject to continuous change, this paper 
begins to address how formal systems for complex systems could be deployed in 
remote teams, to maximize desired ends amidsts constraints and uncertainty.

## Page 23

Instantaneous Remote Teams (IRTs) are generative online-native teams that can 
have rapid evolution of mission, personal composition, skill set, and approach. 
○ Team members are engaged in task allocation, using different practices and 
managing the group’s lifecycle, exchanging results are relevant for the operation 
of the System of Interest (SoI), within the common Markov blanket, using shared 
Ontologies, Narratives, Formal documents, and Tools (ONFT).

## Page 24

As empirical results we want to show some examples of Team_Comm work on this paper. It was 
done by an all-online Remote Team which joined forces to get an overall result which was not 
possible individually (at least in the same timeframe) due to the interdisciplinary nature of the 
research. This team originated unpredictably, following the independent actions of members in a 
shared information niche (Discord channel of the Lex Fridman podcast). Subsequently the team’s 
communication moved to the platform of Keybase which allowed for the construction/development 
of a private informational niche. Several of the topics addressed in this paper can be unpacked 
here in relationship to how we carried out this work: 
● Using Division of Labor 
○ Different members of the Team_Comm have backgrounds in academia, 
Complexity Science, Systems Engineering, and Remote Team management. We 
thus treated the paper as a System of Interest, and through work on its life cycle 
our communication was able to prepare the paper in accordance with best 
research and SE practices. 
● Using ONFT 
○ We were working explicitly with FEP/ActInf and SE Ontologies, figuring out places 
of interconnections/interconnection of concepts from different domains. At online 
weekly meetings we communicated and aligned shared Narratives on different 
levels: about motivation working on these domains, to rise and address questions 
about information we lack, about future application of such approach. We were 
using different tools to support coordination, communication and activities of 
Team_Comm, as well as to create our own information niche and SSoT, based on 
Discord (discord.gg/gpqmQgJ), Keybase channels and subteams (public Keybase 
team @karlfriston.freeenergy, shared Keybase username @ActiveInference). We 
created an external informational presence for Team_Comm activities around 
ActInf which includes Twitter (@InferenceActive) and Youtube (bit.ly/3lg9ztk). 
● Using Alphas 
○ We were training to focus on different aspects in any given time, following the SE 
approach with Alphas for Strategy and Governance in the beginning, later on 
System taking pre-print as an artifact and SoI which states we changed during 
Team_Comm work.

## Page 25

*[Page 25 appears to be blank or image-only]*

## Page 26

Электронный журнал
Том
№

## Page 27

*[Page 27 appears to be blank or image-only]*

## Page 28

*[Page 28 appears to be blank or image-only]*

## Page 29

*[Page 29 appears to be blank or image-only]*

## Page 30

*[Page 30 appears to be blank or image-only]*

## Page 31

*[Page 31 appears to be blank or image-only]*

## Page 32

Инженер и муравей
Блог ШСМ

## Page 33

*[Page 33 appears to be blank or image-only]*

## Page 34

*[Page 34 appears to be blank or image-only]*

## Page 35

*[Page 35 appears to be blank or image-only]*

## Page 36

*[Page 36 appears to be blank or image-only]*


---
*Extraction method: pymupdf*