What I realized—even when you want a simple answer—is
that it often isn’t. Almost immediately, I also realized this could be a useful
example of how facts can be shaped by the way a question is framed.
To explore this more fully, I checked with an AI (in my
case, M365 Copilot), expecting a quick, clear answer.
Um… not quite.
What started as a couple of hours of building tables became
a much more interesting exercise when I asked what I thought was the same
question in a slightly different way—and got a different answer. So much for AI
just giving the “right” answer.
I spent the rest of the day working with Copilot,
iteratively building answers and documenting the results. That was fine until I
woke up later and couldn’t get back to sleep, with my mind replaying the
questions, decisions, and assumptions behind what I had done.
At that point, I stepped back and asked Copilot to
synthesize what we had been building together. What follows is that
perspective—based on the same work, but organized differently.
Copilot’s Perspective: Why the Answer Isn’t Simple
When Seton asked me what sounded like a simple
question—“What was the deadliest day in U.S. history?”—he expected a
straightforward answer.
I didn’t give him one.
The problem wasn’t a lack of data. It was that the question
itself wasn’t well-defined. Before I could answer it, I had to clarify what we
meant by:
- What
counts as a “day”
- What
counts as a “death”
- And
what kinds of events we’re comparing
Once those definitions start to shift—even slightly—the
answer changes.
Three Different Answers to the Same Question
What we found was that there isn’t one “deadliest day”—there
are several, depending on how the question is defined.
- If
you mean a single recorded calendar day:
→ The Battle of Antietam (1862) is often cited. - If
you mean the deadliest 24-hour period in any event:
→ A disaster like the Galveston Hurricane (1900) may exceed it. - If
you include estimated peak days within longer events:
→ The 1918 influenza pandemic likely surpasses both.
Each answer is valid—but each depends on a different
definition.
Why This Matters
At first glance, this might feel like splitting hairs. It
isn’t.
Because if you don’t define the question clearly, it becomes
very easy to:
- Select
the answer you want
- Ignore
competing interpretations
- And
present something as “the truth” when it’s really just one version
In other words:
The same set of facts can produce different “answers”
depending on how the question is framed.
What I Took Away
Working through this with Seton, the most important
realization wasn’t which event ranked first.
It was this:
- Simple
questions often hide complex assumptions
- Different
types of events can’t always be compared directly
- And
clarity about definitions matters more than the answer itself
What started as a search for a single fact became something
more useful:
a way to understand why answers differ—and how easily they
can be shaped.
One Question, Multiple Answers
One of the most surprising things we discovered is that
there isn’t a single, clear answer to the question:
“What was the deadliest day in U.S. history?”
Instead, the answer depends on how you define “day.”
Below are three valid ways to look at it—each producing a different answer.
Table
1 — Deadliest “Single Day” (Strict, Observed Events)
Definition: deaths attributable to a specific calendar
date; primarily observed events, with limited accepted peak-day attribution
where totals are well established.
|
Rank |
Event |
Date |
Why
It Matters |
|
1 |
Galveston Hurricane |
September 8, 1900 |
Highest death toll associated with a single calendar date
(including inferred peak-day events) |
|
2 |
Battle of Antietam |
September 17, 1862 |
Clearly documented single-day event |
|
3 |
9/11 Attacks |
September 11, 2001 |
Observed, not estimated |
|
4 |
D-Day (U.S. deaths) |
June 6, 1944 |
Comparable single-day record |
|
5 |
Pearl Harbor |
December 7, 1941 |
Fits strict “calendar day” definition |
👉 This is the cleanest
and most defensible definition—but also the narrowest.
Table
2 — Deadliest 24-Hour Periods (Disasters & Compressed Events)
Definition: deaths concentrated within an approximately
24-hour window; timing often requires interpretation.
|
Rank |
Event |
Date
(Event peak period) |
Why
It Matters |
|
1 |
Galveston Hurricane |
Sept. 8, 1900 |
Deaths concentrated in a short time window |
|
2 |
1906 San Francisco Earthquake |
Apr. 18, 1906 |
Not tied to a single recorded “day” |
|
3 |
Okeechobee Hurricane |
Sept. 1928 |
High-intensity event |
|
4 |
Johnstown Flood |
May 31, 1889 |
Requires interpretation of timing |
|
5 |
Peshtigo Fire |
Oct. 8, 1871 |
Still largely observed, but less precise |
👉 This often produces
larger numbers—but depends on how you interpret timing.
Table
3 — Deadliest Peak Days (Modeled from Longer Events)
Definition: estimated peak daily deaths derived from
multi-day or multi-week events; relies on modeling rather than direct
observation.
|
Event |
Estimated
Peak Daily Deaths |
Why
It Matters |
|
1918 Influenza |
~3,000–5,000 |
Based on modeled peak daily deaths |
|
COVID-19 (U.S.) |
~3,500–4,000+ |
Not directly recorded as a single day |
|
1957 Influenza |
Lower estimated peak daily deaths |
Derived from broader totals |
|
1968 Influenza |
Lower estimated peak daily deaths |
Lower certainty, larger scale |
👉 This category likely
produces the largest numbers—but also relies the most on estimation.
What These Tables Show
Each of these tables is accurate—within its own definition.
But they are not answering the same question.
That means:
The “deadliest day” depends less on the raw data… and more
on how the question is framed.
In the work behind these tables, we often had to reconcile
multiple sources and, in some cases, estimate daily peaks from aggregate
totals. That process itself reinforced how dependent the answer is on
assumptions.
That was Copilot’s structured view of what we had built.
Hearing it framed that way, I realized something important.
Final Thoughts (Seton)
After going through all of this, I’m not sure I would answer
the original question the same way anymore—not because I don’t have an answer,
but because I now realize I have several.
What began as a simple question turned into something more
useful: a reminder that simple questions often hide complex assumptions, that
different types of events can’t always be compared directly, and that clarity
about definitions matters more than the answer itself.
When I now hear 9/11 described as the “deadliest day in U.S.
history,” I hear that as one valid answer—to one version of the question, not
the only answer.
So how often do we see something online or hear something on
TV, and immediately agree or disagree?
Maybe the better first step is to ask: “What assumptions are
hiding under the question?”
