Want smart analysis of the most important news in your inbox every weekday along with other global reads, interesting ideas and opinions to know? Sign up for the Today’s WorldView newsletter.

It’s no secret that President Trump is desperate to withdraw the United States from the Iran deal. Trump blatantly resents the agreement, the crowning foreign policy achievement of the Obama administration, and lambastes it whenever he can. That included his speech last month at the U.N. General Assembly, during which he called the deal “an embarrassment” and hinted to reporters that he was ready to scrap it.

Now the White House faces its latest deadline — Oct. 15 — to certify to Congress that Iran is in compliance with the terms of the deal. Trump has grudgingly given that approval two times this year, but has loudly signaled that he won’t keep doing so. Speculation over the deal’s demise has given a wing to a flurry of op-eds from neoconservative wonks and Iran hawks in Washington, cheering on possible “decertification.”

But there’s a crucial constituency that may not be on the same page as Trump: “his” generals. On Tuesday, Defense Secretary Jim Mattis, a retired Marine general, appeared before the Senate Armed Services Committee and fielded questions on the administration’s thinking regarding the Iran deal.

During the hearing, Senator Angus King (I-Maine) asked Mattis: “Do you believe it’s in our national security interest at the present time to remain in the JCPOA?” (The JCPOA, or Joint Comprehensive Plan of Action, is the formal term for the Iran deal.)

“Yes, Senator, I do,” Mattis responded.

He also stated that “If we can confirm that Iran is living by the agreement, if we can determine that this is in our best interest, then clearly we should stay with it … I believe at this point in time, absent indications to the contrary, it is something that the president should consider staying with.” Iran’s Press TV pointedly featured the comments on its social media accounts.

US Defense Secretary Mattis backs #Iran nuclear deal pic.twitter.com/OueOgRDzG6

— Press TV (@PressTV) October 3, 2017

Mattis, known for his antipathy toward Iran, is hardly the sort of official whom conservatives can deride as an appeasenik — nor was he the only Marine in the room to back the deal. Gen. Joseph F. Dunford Jr., the chairman of the Joint Chiefs of Staff, said that Iran “is not in material breach” of the deal and that the agreement’s implementation has “delayed the development of nuclear capability by Iran” — precisely as the Obama administration intended.

But the White House sees things differently from the Pentagon. Trump insists that Iran has abrogated the terms of the agreement “in spirit” by testing ballistic missiles and continuing its destabilizing activities in the Middle East. A move by Trump to decertify Iran would give Congress two months to decide whether to restore sanctions on Tehran — an act that could compel Iran to renege on its own commitments. (Congress could also possibly do nothing and let the deal slumber on with European partners taking the lead, which they have pledged to do.)

Another path fervently sought by some Iran hawks is for Congress to use the opportunity to force a renegotiation and expansion of the deal to make it broader than a simple arms control agreement. Critics say the prospect of such a renegotiation is fantasy and, in an interview with Politico’s Susan Glasser, Iranian Foreign Minister Javad Zarif rejected it outright.

“The belief that such an outcome is possible is deeply rooted in a fundamental critique of the Iran deal advanced by some Israeli officials, conservative think tanks, and Trump himself: the idea that Obama simply lacked the skill and political will to push for tougher terms when the JCPOA was struck in July 2015,” wrote Colin Kahl, a former Obama administration official. “There’s just one problem: It’s a myth. More pressure would not have produced a better outcome two years ago — and threatening to blow up the deal will not produce a better one today.”

“The United States will be seen as having violated a deal that was working, the Iranians will feel free to resume their nuclear activities, and the strong international coalition that forced Iran to the table in the first place with its global sanctions will be impossible to put back together,” wrote Philip Gordon of the Council on Foreign Relations in The Washington Post.

Gordon, who also served in the Obama administration, added: “Trying to suddenly conjure up a brand-new deal that addresses every problematic aspect of Iran’s foreign policy and blocks its path to a nuclear weapon is a recipe for achieving neither set of goals.”

Republicans who opposed Iran deal under Obama now think it may be worth saving under Trump, Politico reports:https://t.co/NWMAEbKSwD pic.twitter.com/or2RUMfEi0

— Greg Sargent (@ThePlumLineGS) October 3, 2017

Nevertheless, there’s a real likelihood that Trump will punt the matter to Congress, passing the buck to an already divided and dysfunctional body. “This could very easily take on a momentum of its own,” warned Jon Finer, the chief of staff to one of the Iran deal’s key architects, former secretary of state John F. Kerry. “You could be in a situation where the administration [won’t certify compliance] and Congress will say, ‘Well, the administration has spoken as to Iran’s implementation of the deal,’ and this moves forward whether or not people have fully thought through the consequences.”

One of the consequences of the deal collapsing could be the resumption of Iran’s nuclear program, prompting yet another nuclear-tipped geopolitical crisis. “I know there are some Republicans who are so rabidly against the agreement they might vote to impose sanctions,” said Sen. Chris Murphy (D-Conn.) to Al-Monitor, “but I have to believe there are enough reasonable Republicans that they wouldn’t put both North Korea and Iran on a pathway to nuclear weapons.”

Whatever the mood in Washington, a new poll from the Chicago Council on Global Affairs found that a clear majority of Americans say that the United States should stick to the deal. The country’s top defense officials seem to share that view. But it’s still doubtful that the commander in chief is listening.

Want smart analysis of the most important news in your inbox every weekday along with other global reads, interesting ideas and opinions to know? Sign up for the Today’s WorldView newsletter.

A candlelight vigil is pictured on the Las Vegas strip following a mass shooting at the Route 91 Harvest Country Music Festival in Las Vegas, Nevada, U.S., October 2, 2017. (REUTERS/Chris Wattie TPX IMAGES OF THE DAY)

On Oct. 1, 2017, a man named Stephen Paddock opened fire on attendees at an outdoor concert festival in Las Vegas, killing at least 58 and wounding 500 more, in the deadliest shooting in modern American history.

As the news filtered in, policymakers, pundits and members of the public debated whether the incident should be classified as terrorism. As with other recent violent incidents — including the shootings at Fort Hood and Charleston — opinions diverged widely.

Sheriff Joseph Lombardo of the Las Vegas Metropolitan Police Department announced that the Las Vegas attack was not terrorism, because it was perpetrated by “a solo actor. A lone wolf.”

The FBI made a similar declaration, noting the absence of ties to foreign organizations — although ISIS quickly took credit for the attack regardless. But the Democratic National Committee and Democratic lawmakers like Rep. Jim Cooper (D-Tenn.) characterized the attack as an “act of terror.”

If the perpetrator is Muslim, are we quick to label an incident as “terrorism”?

One question often raised in these incidents is whether violence is more likely to be characterized as terrorism when the perpetrator is a Muslim. On Twitter, journalist Glenn Greenwald wrote that “in the early stages of mass shooting, ‘no signs of terrorism’ means: ‘shooter isn’t Muslim.’”

[Americans who live closer to a mass shooting are more likely to support gun regulation]

In new research, we find that the identity of the perpetrator matters for defining terrorism, but it’s not the only thing that matters. How we define terrorism also depends on other contextual factors that might — or might not — be highlighted in the media.

How do we decide what counts as terrorism?

To help explain this contentious question, we’ve been conducting studies to explore how members of the public decide which violent incidents count as terrorism, and which don’t. In a new article in the American Journal of Political Science, we turn to experimental methods to better understand how the public decides whether violent incidents are acts of terrorism.

If the responses of ordinary citizens constitute a central means through which terrorism operates, understanding what ordinary citizens think counts as terrorism is a crucial prerequisite to understanding how they react to it.

How we did our research

We fielded a survey experiment on 1,400 adult Americans. We showed each respondent summaries of a series of incidents, with randomly generated attributes: the type of tactics used (for example, a shooting versus bombing); if there were casualties; what the target was; where the attack took place; who the perpetrator was (including whether or not they were identified as Muslim); what their motivation was (for example, a personal dispute, policy change, hatred), and so on. We then asked participants to tell us whether they thought of each incident as being an act of terrorism or not. For more on the methodology, see here.

[Lots of Republicans actually support gun control]

Our results showed that ordinary citizens classify terrorism based not only on relatively straightforward facts on the ground, such as the number of casualties — but also information about the perpetrator, which is often highly ambiguous in the immediate aftermath of incidents.

Considerations about the type and severity of violence matter: If the attacker in Las Vegas had used a bomb rather than a gun, for example, we would be much more likely to think of it as terrorism. But our respondents are also heavily influenced by relatively subjective descriptions about the perpetrator’s identity and motivations. These are the “who” and “why” questions — and the media has considerable latitude in how it chooses to answer them, especially in the hours and days following an attack.

Here’s why this matters — predictive models based on our experimental results show that subjective descriptions of the perpetrator make it more/less likely that Americans see the event to be terrorism.

For example, the Daily Mirror speculated about Paddock’s family history of mental illness. Other outlets emphasized other factors. Was he motivated by hatred, or was his motivation unclear? Did he have foreign ties, as the claims by ISIS in the immediate aftermath seemed to suggest? And what if he was Muslim?

Our models suggest that how the public understands an attack depends on which of these factors are mentioned or highlighted. And the media decides which factors make it into the descriptions of attacks that occur.

[America is a violent country]

Here’s the test: building off our experimental results, we calculated the changes in the predicted probability that an event with the characteristics of the Las Vegas attack will be perceived as terrorism if we code the attack’s attributes in different ways. If the perpetrator is described as having a history of mental illness, we have no information about his identity and his motivation is described as unclear, our model suggests a 35 percent likelihood that the public will perceive this as terrorism.

If instead we hold all else equal but describe the perpetrator as Muslim, the probability of viewing it as terrorism increases to 46 percent. If we then omit the references to mental illness, the probability inches up to 53 percent. If we then attribute the incident to political goals, the probability jumps to 73 percent. And, if we also suggest the potential of foreign ties, the probability jumps further to 81 percent.

It isn’t just the attacker’s identity that matters

Tweeting after the attack, Piers Morgan said, “If the shooter was Muslim, we’d call this a terrorist attack.” Our results suggest it’s not that simple: The effect of social identity on terrorism classifications is real (especially among respondents who self-identify as politically conservative), but modest compared to many of the other factors we analyze.

The broader problem is that the media treats Muslim perpetrators differently — they are less likely to be referred to as suffering from histories of mental illness, and their actions more likely to be attributed to political motivations.

These results suggest media coverage profoundly shapes how the public comes to understand violent events. Many of the considerations we rely on most heavily when categorizing events as terrorism are relatively subjective — and require information that’s not available for days or weeks after an incident occurs.

This means pundits and policymakers have considerable leeway in how they choose to frame acts of violence — and our results show that what they report may shape public opinion long before the full details are known.

Connor Huff is a PhD candidate at Harvard University.

Joshua Kertzer is an assistant professor of government at Harvard University.