Originally published on LinkedIn on August 25, 2026.
"Intelligence is whatever intelligence tests measure."
— Edwin Boring, The New Republic, 1923
The previous article ended with a proposition left deliberately open: if organizational co-intelligence is real, it should leave observable traces — the way intelligence in any living system is observable not directly, but through behavior.
This is not a new problem. It is the oldest problem in the science of intelligence. And how researchers solved it — in ecology, in animal cognition, in human psychology — may be the most practically useful thing this series has to offer.
How it is built
In 1997, Suzanne Simard faced a problem familiar to anyone who has tried to study organizational intelligence: the thing she believed existed was entirely invisible.
Her hypothesis — that trees in a British Columbia forest were exchanging carbon through underground fungal networks — had theoretical grounding and observational support. What she could not do was see the network. It existed below ground, inaccessible to direct observation and resistant to every conventional measurement approach.
Her solution was methodologically elegant. She sealed paper birch and Douglas fir seedlings inside plastic chambers, fed one species carbon-14 dioxide and the other carbon-13 dioxide, waited nine days, then analyzed where the labeled carbon had traveled. The result, published in Nature in August 1997, demonstrated bidirectional carbon transfer between the two species through shared mycorrhizal fungal threads.1 The invisible network had not become visible, but it had become traceable — and traceability, it turns out, is enough. If you cannot observe a network directly, you introduce a tracer and follow where it goes.
Organizations appear to exhibit a similar challenge: the mechanisms that allow distributed knowledge to become coordinated action are largely invisible to direct observation. Thomas J. Allen at MIT spent the 1970s measuring a finding whose implications management has never fully absorbed. Across seven research and development laboratories, he documented the relationship between physical distance and the probability of meaningful technical communication,2 and the result was striking in its precision. We are four times as likely to communicate regularly with someone two meters away as with someone twenty meters away. Beyond roughly 50 meters, the relationship flatlines entirely — a different floor of the same building is functionally equivalent, in communication probability, to a different city. The metric system of office planning, Allen found, carries cognitive consequences that almost no one was measuring.
What Allen was measuring was not conversation. He was measuring the conditions under which shared meaning is continuously built — the ambient, informal exchange through which people develop compatible interpretive frameworks over time. Proximity generates the contact through which shared representations form, and those representations are the substrate through which distributed knowledge becomes collectively intelligible. The formal meeting captures what has already been decided; the unplanned hallway exchange is where the interpretive architecture that makes good decisions possible is actually constructed.
A second finding from Allen's research received even less management attention than the distance effect. Across every laboratory he studied, certain individuals served as critical information bridges — conveying concepts between precisely the right people, creating connections the formal organizational chart never showed. These gatekeepers were often poorly recognized by management, sometimes invisible to it entirely. Yet they were the nodes through which the organizational network cohered. When they departed through attrition, restructuring, or neglect, the organization did not simply lose a person. It lost the connections that person had been maintaining for everyone around them.
Subsequent research confirmed that the Allen Curve holds in the digital era. A 2014 study published in the Harvard Business Review found that collocated engineers were 20% more likely to stay in digital contact with each other than those working remotely, and collaborated four times as frequently — producing projects that completed 32% faster.3 The medium of communication had changed; the underlying relationship between proximity and shared meaning proved more durable than anyone had predicted.
MBWA — management by walking around, the practice Hewlett-Packard embedded in its operating culture — was never primarily about visibility or morale. It was network maintenance. The same holds for collocation, regular huddles, all-hands meetings, cross-functional rituals, and offsites. These are the mechanisms through which the organizational meaning network is continuously built and refreshed — investments whose cost is immediately measurable and whose value remains invisible until the network has already degraded, by which point the damage appears elsewhere in symptoms that no longer point to their cause.
How it erodes
In 2022, a research team published a study of 61,182 Microsoft employees in Nature Human Behaviour. The pandemic-imposed shift to remote work had produced a measurable change in organizational network structure: collaboration networks became more static and siloed, cross-group connections dropped by approximately 25% of pre-pandemic levels, and the weak ties through which novel information crosses functional boundaries declined significantly.4
What the study documented was network fragmentation at organizational scale — the individuals still present while the connective tissue between them was thinning invisibly, measurably, and with consequences that would surface later as slower decisions, higher coordination costs, and reduced capacity for the cross-functional integration that organizational intelligence requires.
The erosion of shared meaning tends to be invisible at the moment it occurs. By the time the symptoms appear — misalignment, rework, conflicting priorities, failed initiatives — organizations typically diagnose them as execution or leadership problems. The actual cause goes unexamined because it was never examined to begin with.
How it has always been measured
Boring's 1923 observation was not a definition. It was an accusation: that the nascent science of intelligence had built an elaborate measurement apparatus without establishing what, precisely, it was measuring.
Alfred Binet, who designed the first standardized intelligence test in 1905 with Théodore Simon, understood that measuring intelligence directly was impossible.5 What he could do was identify behavioral proxies — tasks that more intelligent systems perform better than less intelligent ones, consistently enough to be predictive. Those tests did not reveal intelligence so much as trace it, making its presence legible through the behavioral evidence it left behind.
Every subsequent advance in intelligence measurement has followed the same logic, including in non-human systems. We do not ask whether a crow is intelligent. We present it with a novel problem and observe what it does. Frans de Waal spent decades arguing that we had systematically underestimated animal intelligence not because it was absent, but because we had been designing tests that looked for the wrong traces — measuring the intelligence we expected to find rather than the intelligence that was actually there.6
Intelligence has never been directly observed in any system — across a century of research, in humans and animals alike, it has always been inferred from behavioral evidence. The science of intelligence is, in its entirety, a science of traces.
Simard's isotope method was the same epistemological move applied to ecology: find what the invisible network leaves behind and follow where it goes. The organizational equivalent of this move already exists in research. Daniel Wegner's concept of transactive memory — developed in 1987 and later validated in organizational settings by Kyle Lewis — describes the distributed memory architecture that high-performing groups develop over time: individuals know not everything themselves but who in the group knows what, and how to access it quickly.7,8 Transactive memory is measurable, through task performance on novel problems, through network mapping of knowledge retrieval, and through the speed and accuracy with which groups integrate unfamiliar information under pressure. What Wegner and Lewis described is one component of organizational co-intelligence that already has a measurement tradition behind it. Once one component becomes legible, the broader system starts to come into focus.
What organizational co-intelligence leaves behind
If organizational co-intelligence expresses itself through behavior — as all intelligence does — four patterns emerge as its most reliable traces.
How an organization responds to genuine novelty reveals whether its distributed knowledge can integrate into coherent response or fragment into disconnected departmental reactions. Novel situations are to organizational intelligence what unfamiliar problems are to cognitive assessment: they reveal underlying capacity rather than trained response. How quickly relevant information then crosses functional boundaries — and how accurately it is interpreted on arrival — reflects the health of the shared meaning infrastructure, since organizations with degraded shared meaning expend the same energy on information transfer as high-performing ones but with far greater loss in translation.
Whether errors produce changed behavior is the most direct indicator of collective learning. Learning requires not just information about what went wrong but shared understanding of why, distributed to the parts of the system that need to act differently. Without shared meaning, error information circulates without producing change. And how coherently an organization acts under pressure completes the picture — disruption is diagnostic in the same way a behavioral stress test is diagnostic in animal cognition research, making the underlying network topology visible precisely when the system is most strained.
Organizations already generate what is needed to observe all four of these traces. The missing element has never been the evidence but the interpretive framework through which to read it.
What becomes possible
Simard's 1997 research did not only explain forests. It changed how practitioners thought about managing them. Once the network could be demonstrated and partially mapped, the invisible infrastructure that had been routinely destroyed — because it was invisible — became something that could be designed around.
The same transition is available to organizations. Allen's information gatekeepers — individuals who bridge knowledge across groups, often poorly recognized and easily lost in restructuring — become a design consideration rather than an invisible variable once their role in the organizational network can be traced.
Once the infrastructure that maintains shared meaning can be observed and tracked, the investment case changes character entirely — from an intuitive managerial preference to an evidential organizational necessity. The question a CFO asks receives a different answer when the connection between the investment and the organizational output can be demonstrated.
This also changes the job description of a manager more fundamentally than any management framework in recent decades has attempted. The manager whose primary understanding of their role is network maintenance — creating and refreshing the shared meaning through which specialized knowledge remains collectively intelligible — is doing something qualitatively different from coordinating tasks or allocating resources. Those functions matter, and AI can increasingly support them. What AI cannot do is sustain the shared meaning through which tasks acquire coherent organizational purpose, because that maintenance requires the same ambient, proximity-dependent, relationship-sustained contact that the Allen Curve has been measuring for fifty years.
The series began by asking what makes an organization intelligent. Five articles later, the outline of an answer has taken shape. Organizations are cognitive systems that coordinate through shared meaning, build that meaning through proximity and sustained contact, lose it under conditions that modern organizational life creates almost systematically, and leave behavioral traces of both its presence and its absence in data that organizations already collect.
What it takes to build that architecture deliberately — to design for shared meaning rather than leaving it to chance — is where this investigation goes next.
───
Selected References
Empirical Evidence
1. Simard, S. W., Perry, D. A., Jones, M. D., Myrold, D. D., Durall, D. M., & Molina, R. (1997). Net transfer of carbon between ectomycorrhizal tree species in the field.Nature, 388, 579–582.
2. Allen, T. J. (1977). Managing the Flow of Technology. MIT Press.
3. Waber, B., Magnolfi, J., & Lindsay, G. (2014). Workspaces that move people.Harvard Business Review, October 2014, 68–77.
4. Yang, L., Holtz, D., Jaffe, S., et al. (2022). The effects of remote work on collaboration among information workers.Nature Human Behaviour, 6, 43–54.
Foundations of Intelligence and Organizational Research
5. Binet, A., & Simon, T. (1905). Méthodes nouvelles pour le diagnostic du niveau intellectuel des anormaux.L'Année Psychologique, 11, 191–244.
6. de Waal, F. (2016). Are We Smart Enough to Know How Smart Animals Are? W.W. Norton.
7. Wegner, D. M. (1987). Transactive memory: A contemporary analysis of the group mind. In B. Mullen & G. R. Goethals (Eds.), Theories of Group Behavior. Springer.
8. Lewis, K. (2003). Measuring transactive memory systems in the field: Scale development and validation. Journal of Applied Psychology, 88(4), 587–604.
9. Peters, T. J., & Waterman, R. H. (1982). In Search of Excellence. Harper & Row.
10. Gardner, H. (1983). Frames of Mind: The Theory of Multiple Intelligences. Basic Books.
11. Boring, E. G. (1923). Intelligence as the tests test it. The New Republic, 36, 35–37.
Explore the OCI series
Previous: The Illusion of Alignment
Next: The Infrastructure Nobody Designed
Discuss organizational co-intelligence in your organization.
Originally published on LinkedIn on August 25, 2026.
"Intelligence is whatever intelligence tests measure."
— Edwin Boring, The New Republic, 1923
The previous article ended with a proposition left deliberately open: if organizational co-intelligence is real, it should leave observable traces — the way intelligence in any living system is observable not directly, but through behavior.
This is not a new problem. It is the oldest problem in the science of intelligence. And how researchers solved it — in ecology, in animal cognition, in human psychology — may be the most practically useful thing this series has to offer.
How it is built
In 1997, Suzanne Simard faced a problem familiar to anyone who has tried to study organizational intelligence: the thing she believed existed was entirely invisible.
Her hypothesis — that trees in a British Columbia forest were exchanging carbon through underground fungal networks — had theoretical grounding and observational support. What she could not do was see the network. It existed below ground, inaccessible to direct observation and resistant to every conventional measurement approach.
Her solution was methodologically elegant. She sealed paper birch and Douglas fir seedlings inside plastic chambers, fed one species carbon-14 dioxide and the other carbon-13 dioxide, waited nine days, then analyzed where the labeled carbon had traveled. The result, published in Nature in August 1997, demonstrated bidirectional carbon transfer between the two species through shared mycorrhizal fungal threads.1 The invisible network had not become visible, but it had become traceable — and traceability, it turns out, is enough. If you cannot observe a network directly, you introduce a tracer and follow where it goes.
Organizations appear to exhibit a similar challenge: the mechanisms that allow distributed knowledge to become coordinated action are largely invisible to direct observation. Thomas J. Allen at MIT spent the 1970s measuring a finding whose implications management has never fully absorbed. Across seven research and development laboratories, he documented the relationship between physical distance and the probability of meaningful technical communication,2 and the result was striking in its precision. We are four times as likely to communicate regularly with someone two meters away as with someone twenty meters away. Beyond roughly 50 meters, the relationship flatlines entirely — a different floor of the same building is functionally equivalent, in communication probability, to a different city. The metric system of office planning, Allen found, carries cognitive consequences that almost no one was measuring.
What Allen was measuring was not conversation. He was measuring the conditions under which shared meaning is continuously built — the ambient, informal exchange through which people develop compatible interpretive frameworks over time. Proximity generates the contact through which shared representations form, and those representations are the substrate through which distributed knowledge becomes collectively intelligible. The formal meeting captures what has already been decided; the unplanned hallway exchange is where the interpretive architecture that makes good decisions possible is actually constructed.
A second finding from Allen's research received even less management attention than the distance effect. Across every laboratory he studied, certain individuals served as critical information bridges — conveying concepts between precisely the right people, creating connections the formal organizational chart never showed. These gatekeepers were often poorly recognized by management, sometimes invisible to it entirely. Yet they were the nodes through which the organizational network cohered. When they departed through attrition, restructuring, or neglect, the organization did not simply lose a person. It lost the connections that person had been maintaining for everyone around them.
Subsequent research confirmed that the Allen Curve holds in the digital era. A 2014 study published in the Harvard Business Review found that collocated engineers were 20% more likely to stay in digital contact with each other than those working remotely, and collaborated four times as frequently — producing projects that completed 32% faster.3 The medium of communication had changed; the underlying relationship between proximity and shared meaning proved more durable than anyone had predicted.
MBWA — management by walking around, the practice Hewlett-Packard embedded in its operating culture — was never primarily about visibility or morale. It was network maintenance. The same holds for collocation, regular huddles, all-hands meetings, cross-functional rituals, and offsites. These are the mechanisms through which the organizational meaning network is continuously built and refreshed — investments whose cost is immediately measurable and whose value remains invisible until the network has already degraded, by which point the damage appears elsewhere in symptoms that no longer point to their cause.
How it erodes
In 2022, a research team published a study of 61,182 Microsoft employees in Nature Human Behaviour. The pandemic-imposed shift to remote work had produced a measurable change in organizational network structure: collaboration networks became more static and siloed, cross-group connections dropped by approximately 25% of pre-pandemic levels, and the weak ties through which novel information crosses functional boundaries declined significantly.4
What the study documented was network fragmentation at organizational scale — the individuals still present while the connective tissue between them was thinning invisibly, measurably, and with consequences that would surface later as slower decisions, higher coordination costs, and reduced capacity for the cross-functional integration that organizational intelligence requires.
The erosion of shared meaning tends to be invisible at the moment it occurs. By the time the symptoms appear — misalignment, rework, conflicting priorities, failed initiatives — organizations typically diagnose them as execution or leadership problems. The actual cause goes unexamined because it was never examined to begin with.
How it has always been measured
Boring's 1923 observation was not a definition. It was an accusation: that the nascent science of intelligence had built an elaborate measurement apparatus without establishing what, precisely, it was measuring.
Alfred Binet, who designed the first standardized intelligence test in 1905 with Théodore Simon, understood that measuring intelligence directly was impossible.5 What he could do was identify behavioral proxies — tasks that more intelligent systems perform better than less intelligent ones, consistently enough to be predictive. Those tests did not reveal intelligence so much as trace it, making its presence legible through the behavioral evidence it left behind.
Every subsequent advance in intelligence measurement has followed the same logic, including in non-human systems. We do not ask whether a crow is intelligent. We present it with a novel problem and observe what it does. Frans de Waal spent decades arguing that we had systematically underestimated animal intelligence not because it was absent, but because we had been designing tests that looked for the wrong traces — measuring the intelligence we expected to find rather than the intelligence that was actually there.6
Intelligence has never been directly observed in any system — across a century of research, in humans and animals alike, it has always been inferred from behavioral evidence. The science of intelligence is, in its entirety, a science of traces.
Simard's isotope method was the same epistemological move applied to ecology: find what the invisible network leaves behind and follow where it goes. The organizational equivalent of this move already exists in research. Daniel Wegner's concept of transactive memory — developed in 1987 and later validated in organizational settings by Kyle Lewis — describes the distributed memory architecture that high-performing groups develop over time: individuals know not everything themselves but who in the group knows what, and how to access it quickly.7,8 Transactive memory is measurable, through task performance on novel problems, through network mapping of knowledge retrieval, and through the speed and accuracy with which groups integrate unfamiliar information under pressure. What Wegner and Lewis described is one component of organizational co-intelligence that already has a measurement tradition behind it. Once one component becomes legible, the broader system starts to come into focus.
What organizational co-intelligence leaves behind
If organizational co-intelligence expresses itself through behavior — as all intelligence does — four patterns emerge as its most reliable traces.
How an organization responds to genuine novelty reveals whether its distributed knowledge can integrate into coherent response or fragment into disconnected departmental reactions. Novel situations are to organizational intelligence what unfamiliar problems are to cognitive assessment: they reveal underlying capacity rather than trained response. How quickly relevant information then crosses functional boundaries — and how accurately it is interpreted on arrival — reflects the health of the shared meaning infrastructure, since organizations with degraded shared meaning expend the same energy on information transfer as high-performing ones but with far greater loss in translation.
Whether errors produce changed behavior is the most direct indicator of collective learning. Learning requires not just information about what went wrong but shared understanding of why, distributed to the parts of the system that need to act differently. Without shared meaning, error information circulates without producing change. And how coherently an organization acts under pressure completes the picture — disruption is diagnostic in the same way a behavioral stress test is diagnostic in animal cognition research, making the underlying network topology visible precisely when the system is most strained.
Organizations already generate what is needed to observe all four of these traces. The missing element has never been the evidence but the interpretive framework through which to read it.
What becomes possible
Simard's 1997 research did not only explain forests. It changed how practitioners thought about managing them. Once the network could be demonstrated and partially mapped, the invisible infrastructure that had been routinely destroyed — because it was invisible — became something that could be designed around.
The same transition is available to organizations. Allen's information gatekeepers — individuals who bridge knowledge across groups, often poorly recognized and easily lost in restructuring — become a design consideration rather than an invisible variable once their role in the organizational network can be traced.
Once the infrastructure that maintains shared meaning can be observed and tracked, the investment case changes character entirely — from an intuitive managerial preference to an evidential organizational necessity. The question a CFO asks receives a different answer when the connection between the investment and the organizational output can be demonstrated.
This also changes the job description of a manager more fundamentally than any management framework in recent decades has attempted. The manager whose primary understanding of their role is network maintenance — creating and refreshing the shared meaning through which specialized knowledge remains collectively intelligible — is doing something qualitatively different from coordinating tasks or allocating resources. Those functions matter, and AI can increasingly support them. What AI cannot do is sustain the shared meaning through which tasks acquire coherent organizational purpose, because that maintenance requires the same ambient, proximity-dependent, relationship-sustained contact that the Allen Curve has been measuring for fifty years.
The series began by asking what makes an organization intelligent. Five articles later, the outline of an answer has taken shape. Organizations are cognitive systems that coordinate through shared meaning, build that meaning through proximity and sustained contact, lose it under conditions that modern organizational life creates almost systematically, and leave behavioral traces of both its presence and its absence in data that organizations already collect.
What it takes to build that architecture deliberately — to design for shared meaning rather than leaving it to chance — is where this investigation goes next.
───
Selected References
Empirical Evidence
1. Simard, S. W., Perry, D. A., Jones, M. D., Myrold, D. D., Durall, D. M., & Molina, R. (1997). Net transfer of carbon between ectomycorrhizal tree species in the field.Nature, 388, 579–582.
2. Allen, T. J. (1977). Managing the Flow of Technology. MIT Press.
3. Waber, B., Magnolfi, J., & Lindsay, G. (2014). Workspaces that move people.Harvard Business Review, October 2014, 68–77.
4. Yang, L., Holtz, D., Jaffe, S., et al. (2022). The effects of remote work on collaboration among information workers.Nature Human Behaviour, 6, 43–54.
Foundations of Intelligence and Organizational Research
5. Binet, A., & Simon, T. (1905). Méthodes nouvelles pour le diagnostic du niveau intellectuel des anormaux.L'Année Psychologique, 11, 191–244.
6. de Waal, F. (2016). Are We Smart Enough to Know How Smart Animals Are? W.W. Norton.
7. Wegner, D. M. (1987). Transactive memory: A contemporary analysis of the group mind. In B. Mullen & G. R. Goethals (Eds.), Theories of Group Behavior. Springer.
8. Lewis, K. (2003). Measuring transactive memory systems in the field: Scale development and validation. Journal of Applied Psychology, 88(4), 587–604.
9. Peters, T. J., & Waterman, R. H. (1982). In Search of Excellence. Harper & Row.
10. Gardner, H. (1983). Frames of Mind: The Theory of Multiple Intelligences. Basic Books.
11. Boring, E. G. (1923). Intelligence as the tests test it. The New Republic, 36, 35–37.
Explore the OCI series
Previous: The Illusion of Alignment
Next: The Infrastructure Nobody Designed
Discuss organizational co-intelligence in your organization.