So, if data can reveal these hidden patterns, what responsibility do we have to act on them before the next tragedy occurs?

Even with limited data, the model has achieved measurable accuracy, identifying high-risk periods months in advance. It proves that what we’ve long considered “rare” may simply be what we haven’t yet learned to recognize.

At the Lewiston Institute, we took that challenge seriously. Our system, SentinelForecast360, uses patterns found in environmental and temporal factors — not human profiling — to forecast when and where mass shootings are most likely to occur.

The National Institute of Justice has long noted that the real challenge isn’t a lack of desire to understand these tragedies, but a lack of consistent data. Explaining, “The research community must develop a strategy to source, code, check, and analyze the data surrounding rare violent incidents.”

Experts describe mass shootings as “rare events” — the kind that data scientists call outliers, and policymakers label “impossible to foresee.”

Closing Question: If predictive modeling can be designed to protect lives without bias, shouldn’t we explore how to use it responsibly?

By removing demographic and socioeconomic variables, SentinelForecast360 avoids the pitfalls of biased policing models and instead provides ethical, data-driven foresight designed to protect communities, not profile them.

It uses objective, non-identifiable data such as environmental conditions, temporal patterns, and historical incidents to identify when and where risk is statistically highest.

At the Lewiston Institute, our approach is different. SentinelForecast360 focuses exclusively on forecasting mass shooting risks—not individual behavior.

The Brennan Center for Justice cautions that many predictive policing algorithms amplify existing systemic biases, noting that these systems “use algorithms to analyze massive amounts of information in order to predict and help prevent potential future crimes.”

You’re absolutely right that predictive technology can be dangerous in the wrong hands. But when built ethically and transparently, it can also empower communities and policymakers to act before lives are lost — without violating anyone’s civil rights.

Unlike mass surveillance systems or “pre-crime” units, our model doesn’t monitor individuals or use personal data. It analyzes environmental and behavioral patterns drawn from open-source incident data to identify where conditions for violence are statistically forming — not who will commit it.

You raise a valid point — much of today’s AI is profit-driven, designed to serve corporate or political interests rather than public safety or human rights. That’s one of the biggest reasons I built SentinelForecast360 independently.

6️⃣ Closing Question: If AI can detect patterns that signal risk before tragedy strikes, should we not use it to save lives?

5️⃣ This shows that while AI may not “think” like humans, it can reveal patterns invisible to us, giving communities and law enforcement actionable foresight.

4️⃣ In 2025, SentinelForecast360 identified April and August as the highest-risk months for mass shootings in Los Angeles—and eight tracked incidents confirmed the forecast, achieving 76 percent predictive accuracy.

3️⃣ At the Lewiston Institute, we challenged that assumption. Our AI system, SentinelForecast360, doesn’t try to predict human intent—it analyzes historical incidents, environmental conditions, and temporal patterns to forecast risk.

2️⃣ Deutsch says: “a mere AI is incapable of having any such ideas, because the capacity for considering them has been designed out of its constitution.”

It'll be messy!