Why complex transitions may not have a single date — and why that matters before we try to predict what happens next
In my previous article, The Real Question for Canada Isn’t How to Beat America, I introduced a developing Tang Papers idea called the Representational Transition Signature, or RTS.
It emerged from a simple problem.
Canada was clearly moving.
Its negotiating behaviour was changing. Governments were talking about diversification. Businesses were reconsidering supply chains. New trade relationships and infrastructure were being discussed.
But movement, I argued, is not the same as transition.
A minister visiting another country is movement.
A company announcing that it wants another supplier is movement.
A new trade mission is movement.
None of those things, by itself, proves that the underlying structure of the Canada–U.S. relationship has changed.
That led me to a question:
When did the transition actually happen?
RTS deliberately refuses to assume that there must be one answer.
Politics might change on one date.
Negotiating behaviour on another.
Trade patterns later.
Infrastructure later still.
Institutions may not change at all.
And sometimes the evidence simply isn’t sufficient to tell us.
In the earlier article, I reduced the idea to three deliberately simple possibilities:
Changed.
Didn’t change.
Or we don’t yet know.
At first, I thought this was mainly a problem of identifying and dating change.
But that question has since led me somewhere more interesting.
Because once we ask when a transition happened, another question appears:
What were we actually entitled to say had changed before we knew how the story ended?
And then comes the practical problem:
What if we have to make a decision before we know the answer?
History Knows the Ending
Imagine a company that collapses five years from now.
Looking backward, the turning point may seem obvious.
Revenue had started slowing.
A major customer had left.
Complaints were increasing.
A competitor had introduced something new.
An important employee had departed.
Five years later, we can arrange those events into a convincing story.
There were the warning signs.
There was the beginning of the decline.
There was the moment everything changed.
But now imagine standing inside the company when those things were actually happening.
Was the company failing? Or was it having a difficult quarter?
Was the lost customer the beginning of a structural problem or simply one lost customer?
Was the employee’s departure significant?
Would sales recover?
At the time, several futures were still possible.
Knowing the ending changes our position as observers.
This isn’t a new discovery.
In 1975, psychologist Baruch Fischhoff published his classic research on what became known as hindsight bias.
His experiments showed that once people knew how something had turned out, that knowledge could change how they evaluated what came before. Outcomes that had once been uncertain could begin to appear more predictable than they really were.
Decades of subsequent research have explored the same general problem in areas ranging from medicine and law to historical analysis and accident investigation.
So RTS does not propose that hindsight bias is new.
It raises a narrower question:
What if we preserve our representation of a transition before hindsight gets the opportunity to rewrite it?
A Record Before a Signature
The name Representational Transition Signature carries an implication that needs to be earned.
A signature suggests that recurring structure exists.
We don’t know that yet.
At this stage, the more modest object is really a transition record.
Suppose we’re watching several distinct parts of a system.
One has clearly changed.
Another hasn’t changed within the period we’re observing.
A third remains unresolved.
Instead of collapsing them into:
“The system has transitioned.”
we preserve the differences:
Changed.
Didn’t change.
Don’t know.
The purpose isn’t to make uncertainty sound sophisticated.
It is to preserve what the evidence actually justified at that moment.
Then freeze the record.
Don’t quietly rewrite it when the ending becomes known.
Later, construct another account using the evidence now available.
Now we have something interesting to compare.
What did the transition look like while we were inside it?
What does it look like after we know what happened?
The difference between those two representations may itself contain information.
What Does Hindsight Change?
Suppose our original record said:
Economic structure: Don’t know.
Five years later, everyone agrees that the economy underwent a major structural transition.
It becomes tempting to go back and say:
“Clearly the transition was already underway.”
Perhaps it was.
But that isn’t quite the same statement.
One describes what we can see afterward.
The other describes what the evidence justified beforehand.
Those positions should not be confused.
The same problem can occur with events that never develop into transitions.
Something looks enormously consequential while it is happening.
Five years later, it barely appears in the historical narrative.
So hindsight can potentially work in both directions.
It can make an eventual turning point look inevitable.
And it can make unresolved pathways that didn’t occur disappear from the story.
A prospective record preserves something the retrospective account cannot fully reconstruct:
the uncertainty that actually existed at the time.
That creates an empirical question for RTS.
If we repeatedly preserve records before outcomes are known, then later compare them with retrospective reconstructions, do systematic differences appear?
Does hindsight tend to move transition dates earlier?
Does it eliminate “don’t know”?
Does it make sequences look cleaner?
Does it make one event appear more causally important?
Does it compress several different transitions into one apparent turning point?
In other words, does knowing the outcome systematically simplify the representation of the transition that preceded it?
I don’t know.
But that is now a question we can test.
But We Don’t Get to Wait for History
This is where the problem becomes much more practical.
Canada cannot wait twenty years for historians to decide whether 2026 represented a temporary disruption or a structural transformation of the Canada–U.S. relationship.
Businesses cannot wait.
Investors cannot wait.
Governments cannot wait.
They have to make decisions while the transition remains unresolved.
The struggling company in our earlier example has exactly the same problem.
If management decides too quickly that its business model has fundamentally changed, it might restructure the company, abandon good products or spend heavily entering new markets.
But suppose the disruption was temporary.
It may have constructed a solution to the wrong problem.
Now reverse it.
Management assumes everything is temporary.
It waits for conditions to return to normal.
But the underlying market really has changed.
Now the company may be optimizing itself for a world that is disappearing.
The difficulty isn’t merely that we can predict the future incorrectly.
There’s a problem that comes before prediction:
We can represent the present incorrectly.
Before “What Happens Next?”
We spend enormous effort asking:
What will markets do?
What will another country do?
What will artificial intelligence change?
Where will interest rates go?
What will consumers want?
Those are questions about the future.
But every forecast starts somewhere.
Before asking:
What happens next?
we have already made assumptions about:
What is happening now?
That’s where RTS potentially fits.
RTS is not itself a forecasting model.
Its first job is representational.
It asks:
What are we actually justified in treating as having changed?
A separate question is whether preserving that structure eventually improves forecasting compared with simpler ways of representing the same situation.
That has to be demonstrated, not assumed.
“Don’t Know” Doesn’t Mean “Do Nothing”
There is an obvious danger in emphasizing uncertainty.
It can sound like an argument for paralysis.
If we don’t know whether a transition has occurred, perhaps we should simply wait.
But real decision-makers often don’t have that luxury.
Uncertainty doesn’t necessarily tell us to do nothing.
It should constrain what assumptions we build into what we do.
Suppose two futures remain plausible.
One strategy works brilliantly if Future A occurs but fails badly under Future B.
Another strategy performs reasonably well under both.
Which should we choose?
There is already a substantial body of research addressing problems like this.
Robust Decision Making, developed by Robert Lempert and colleagues at RAND, uses models and data to stress-test proposed strategies across many plausible futures rather than relying on a single prediction of what the future will be.
RTS should not compete with or attempt to replace that work.
Its possible role comes earlier.
Robust decision methods can ask:
What should we do given uncertainty about the future?
RTS asks:
What are we justified in treating as having already changed when we construct that decision problem?
That distinction matters.
Don’t Remove Uncertainty Before the World Does
Suppose part of a system has clearly transitioned.
Another part hasn’t.
A third is moving.
A fourth remains genuinely unresolved.
We could summarize all of that by saying:
“The system is transitioning.”
That may sometimes be perfectly adequate.
But perhaps we’ve also thrown something away.
The unresolved structure may itself contain information.
That leads to a principle that has emerged from this exploration:
Don’t eliminate uncertainty in the representation before the evidence eliminates it in the world.
And when that representation is passed into a decision process, the practical consequence follows:
Don’t build a solution that requires the uncertainty to have disappeared when it hasn’t.
Neither statement means uncertainty should be preserved forever.
New evidence should resolve uncertainty when it can.
Nor are these claims that RTS has invented decision-making under uncertainty. It hasn’t.
The narrower question is whether preserving which parts of a transition have changed, which haven’t, and which remain unresolved provides useful information that is lost when the whole system is compressed into a simpler state.
That is something RTS would have to earn empirically.
Back to Canada
This brings us back to the case that produced the question.
Has the Canada–U.S. relationship transitioned?
The most responsible answer remains:
It depends on what part of the relationship we’re talking about.
Some things have clearly moved.
Some behaviour has changed.
Some underlying structures remain deeply embedded.
Some alternatives are being developed but are not yet usable at sufficient scale.
And some outcomes remain unknown.
Canada nevertheless has to make decisions about trade, infrastructure, energy, defence, industrial capacity and economic relationships while that uncertainty remains.
There are two symmetrical dangers.
Canada could assume too quickly that the old relationship is gone and construct expensive solutions around a transition that proves less permanent than expected.
Or it could assume that the previous relationship will return and remain dependent on pathways whose reliability is already changing.
Perhaps the immediate objective therefore isn’t perfect prediction.
It may be to ask:
What can Canada build now that remains useful across several plausible things the United States might do next?
That isn’t an RTS solution. It is a strategy question.
But the representation supplied to that strategy process matters.
Before deciding what Canada should build, we should be explicit about:
What has changed?
What hasn’t?
What remains unresolved?
Then strategy can work with that uncertainty rather than quietly assuming it away.
David Has the Same Problem
This also connects with another question I have been exploring:
What if David doesn’t have to beat Goliath?
Perhaps David remains smaller while becoming harder for Goliath to constrain.
But how do we know when that has happened?
David announces another trade relationship.
Has the constraint disappeared?
Not necessarily.
David finds another supplier.
Is the alternative usable at sufficient scale?
Not necessarily.
David starts constructing another pathway.
Has the transition occurred?
Not necessarily.
As I argued in the earlier Canada essay:
An alternative isn’t an option merely because you can see it.
And:
Movement is not transition.
David therefore has two different problems.
First:
What has actually changed?
Then:
Given what has changed—and what remains unresolved—what should I do next?
RTS is concerned primarily with the first.
Forecasting, scenario planning, robust decision-making and strategy tackle versions of the second.
The interesting possibility is whether a more disciplined answer to the first improves what we can do with the second.
Think About Artificial Intelligence
Consider how often we hear:
“AI has changed everything.”
Perhaps.
But what exactly has transitioned?
Technical capability?
Adoption?
Workflows?
Employment?
Productivity?
Education?
Regulation?
These things are related, but they don’t necessarily move together.
Technical capability can change rapidly.
Adoption can be uneven.
Organizations can change workflows at different speeds.
Employment effects may emerge differently across industries.
Productivity may take longer to measure.
Education and regulation can move on still other clocks.
Yet companies, workers, schools and governments have to make decisions now.
Compress everything into:
“The AI transition has happened.”
and we may behave as though uncertainties have already resolved.
Dismiss everything as hype, and we may make the opposite mistake.
The more useful representation might be messier:
This changed.
This hasn’t.
This is moving.
This remains unresolved.
Then ask what actions remain sensible given that combination.
The Systems Don’t Have to Be the Same
There is an important boundary here.
Canada–U.S. relations are not artificial intelligence.
AI is not a struggling company.
A company is not a marriage.
Finding a similar representational problem across those examples does not mean they operate through the same mechanisms.
The cross-domain possibility is narrower.
Different systems may produce similar errors in how observers represent change, even when the mechanisms producing those changes are completely different.
We take several clocks and create one date.
We take movement and call it transition.
We observe sequence and infer causality.
We see an alternative and assume it is usable.
We know the ending and reconstruct the beginning.
Perhaps the recurring feature isn’t the system.
Perhaps it’s the observer.
That possibility connects RTS much more directly with the broader Tang Papers research program.
The underlying concern isn’t simply transition.
It’s representation.
What happens when complicated reality is compressed into something simpler?
And what gets lost?
RTS Has to Compete With Simpler Answers
There is an uncomfortable possibility that needs to remain visible.
Perhaps RTS isn’t necessary.
A more complicated representation is not automatically a better representation.
Maybe recording several domain-level transitions preserves useful information.
Or perhaps conventional ways of representing uncertainty already capture everything that matters with less complexity.
If so, we should prefer the simpler method.
So the research question cannot be:
Can we find examples that fit RTS?
We almost certainly can.
The harder question is:
Does preserving independently justified differences among parts of an unfolding transition provide information that is lost when the system is represented more simply?
And then:
Does that information matter for anything downstream?
For understanding?
For forecasting?
For decisions?
Those questions allow RTS to fail.
They need to.
A research framework becomes more useful when we know what evidence could make it unnecessary.
A Prospective Experiment
Canada gives us an unusual opportunity. We are still inside the story.
That means we can preserve our interpretation now.
Not five years from now.
Not after we know how the confrontation ends.
Now.
We can record what appears to have changed.
What hasn’t.
What remains unresolved.
Then leave the record alone.
Later, once more of the outcome is known, construct a new account independently.
Now compare them.
Did hindsight move the apparent transition dates?
Did uncertainty disappear from the retrospective story?
Did events that seemed important at the time disappear?
Did events that eventually mattered begin to look more obvious than they really were?
Did knowing the outcome simplify the apparent structure of the transition?
And most importantly for the longer research program: Did preserving the original transition structure provide information that proved useful later?
Canada alone cannot validate RTS.
It is not a control group.
It can, however, become a prospectively frozen observational case rather than another historical example chosen after we already know the answer.
That difference matters.
The Record Is Not the Prediction
There is a final distinction worth preserving.
A prospective RTS record may be wrong.
Something classified as unresolved today may become obvious tomorrow.
Something that appears to have changed may reverse.
The purpose isn’t to claim that the observer standing inside history understands the system better than the historian looking backward.
The historian will often know much more.
The purpose is to prevent the two observational positions from becoming indistinguishable.
RTS asks:
What did the evidence justify at the time?
Forecasting asks:
What might happen next?
Decision analysis asks:
What should we do given the possible futures?
Those are different questions.
But they may form a useful sequence:
What changed?
What didn’t?
What remains unresolved?
What might happen next?
What can we do without pretending to know which future will occur?
This sequence should not be mistaken for a new universal Tang Papers architecture.
It is simply a way of keeping different analytical tasks from collapsing into one another.
Before History Tells Us the Answer
In the earlier Canada–U.S. article, I ended with a discipline that continues to feel important:
Don’t finish the story before the evidence does.
RTS takes that discipline one step further.
If we’re going to study transitions prospectively, preserve what we knew before the story ended.
Changed.
Didn’t change.
Don’t know.
Then let reality continue.
Later, compare what we thought was happening with what became visible afterward.
Perhaps the difference tells us something about hindsight.
Perhaps it tells us something about transition.
Perhaps preserving that structure will eventually improve forecasting or decision-making.
Or perhaps established, simpler representations of uncertainty will perform just as well.
All of those are legitimate outcomes.
Because the deeper question is no longer simply:
When did the transition actually happen?
It is:
Can we preserve a multidimensional record of change before the ending is known—and does preserving it eventually help us understand, predict or act better than if we had compressed the transition too soon?
We don’t know yet.
And that is precisely why the record has to be made now.
Preserve the evidence before compressing the story.
And when decisions cannot wait:
Don’t build a solution that requires the uncertainty to have disappeared when it hasn’t.
We spend enormous effort trying to understand the world we’re moving toward.
Sometimes the harder problem comes first.
Understanding the world we’re actually standing in now.
About the Author
Robert (Lit Meng) Tang is an independent researcher, writer and entrepreneur based in Ontario, Canada. He holds a B.Sc. in Mathematics from McMaster University and an MBA from the Schulich School of Business at York University.
His professional background spans research, marketing and strategic communications. He previously served as Research Manager with the Society of Management Accountants of Canada and later co-founded danceScape with his wife and business partner, Beverley Cayton-Tang.
Tang is the author of the Tang Papers, an independent research program exploring representation, relational structure, information, coordination and transition across complex systems.
This essay continues the exploration begun in The Real Question for Canada Isn’t How to Beat America, where the Representational Transition Signature (RTS) was introduced publicly.
RTS remains a developing research framework. It is not a validated forecasting or decision-making model. Its emerging research question is whether preserving independently warranted differences among domain-level transitions—including transition, non-transition and unresolved states—retains useful information that simpler representations discard, and whether any such information demonstrates downstream value when tested prospectively.
The larger Tang Papers research program is available through the danceScape Research Archive.
References
Fischhoff, B. (1975). Hindsight is not equal to foresight: The effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance, 1(3), 288–299. DOI: 10.1037/0096-1523.1.3.288.
Fischhoff, B. (2025). Fifty years of hindsight bias research—Reflection on Fischhoff (1975). Journal of Experimental Psychology: Human Perception and Performance, 51(2), 143–150. DOI: 10.1037/xhp0001232.
Lempert, R. J. (2019). Robust Decision Making (RDM). In V. A. W. J. Marchau, W. E. Walker, P. J. T. M. Bloemen, & S. W. Popper (Eds.), Decision Making under Deep Uncertainty: From Theory to Practice (pp. 23–51). Springer, Cham. DOI: 10.1007/978-3-030-05252-2_2.