Ming Xie, University of Maryland, Baltimore County

Imagine a world in which a hurricane devastates the Gulf Coast, and the U.S. has no federal agency prepared to quickly send supplies, financial aid and temporary housing assistance.

Could the states manage this catastrophic event on their own?

Normally, the Federal Emergency Management Agency, known as FEMA, is prepared to marshal supplies within hours of a disaster and begin distributing financial aid to residents who need help.

However, with President Donald Trump questioning FEMA’s future and suggesting states take over recovery instead, and climate change causing more frequent and severe disasters, it’s worth asking how prepared states are to face these growing challenges without help.

What FEMA does

FEMA was created in 1979 with the job of coordinating national responses to disasters, but the federal government has played important roles in disaster relief since the 1800s.

During a disaster, FEMA’s assistance can begin only after a state requests an emergency declaration and the U.S. president approves it. The request has to show that the disaster is so severe that the state can’t handle the response on its own.

FEMA’s role is to support state and local governments by coordinating federal agencies and providing financial aid and recovery assistance that states would otherwise struggle to supply on their own. FEMA doesn’t “take over,” as a misinformation campaign launched during Hurricane Helene claimed. Instead, it pools federal resources to allow states to recover faster from expensive disasters.

During a disaster, FEMA:

• Coordinates federal resources. For example, during Hurricane Ian in 2022, FEMA coordinated with the U.S. Coast Guard, the Department of Defense and search-and-rescue teams to conduct rescue operations, organized utility crews to begin restoring power and also delivered water and millions of meals.

• Provides financial assistance. FEMA distributes billions of dollars in disaster relief funds to help individuals, businesses and local governments recover. As of Feb. 3, 2025, FEMA aid from 2024 storms included US$1.04 billion related to Hurricane Milton, $416.1 million for Hurricane Helene and $112.6 million for Hurricane Debby.

• Provides logistical support. FEMA coordinates with state and local governments, nonprofits such as the American Red Cross and federal agencies to supply cots, blankets and hygiene supplies for emergency shelters. It also works with state and local partners to distribute critical supplies such as food, water and medical aid.

The agency also manages the National Flood Insurance Program, offers disaster preparedness training and helps states develop response plans to improve their overall responses systems.

What FEMA aid looks like in a disaster

When wildfires swept through Maui, Hawaii, in August 2023, FEMA provided emergency grants to cover immediate needs such as food, clothing and essential supplies for survivors.

The agency arranged hotel rooms, rental assistance and financial aid for residents who lost homes or belongings. Its Direct Housing Program has spent $295 million to lease homes for more than 1,200 households. This comprehensive support helped thousands of people begin rebuilding their lives after losing almost everything.

FEMA also helped fund construction of a temporary school to ensure that students whose schools burned could continue their classes. Hawaii, with its relatively small population and limited emergency funds, would have struggled to mount a comparable response on its own.

Larger states often need help, too. When a 2021 winter storm overwhelmed Texas’ power grid and water infrastructure, FEMA coordinated the delivery of essential supplies, including water, fuel, generators and blankets, following the disaster declaration on Feb. 19, 2021. Within days, it awarded more than $2.8 million in grants to help people with temporary housing and home repairs.

Which states would suffer most without FEMA?

Without FEMA or other federal support, states would have to manage the disaster response and recovery on their own.

States prone to frequent disasters, such as Louisiana and Florida, would face expensive recurring challenges that would likely exacerbate recovery delays and reduce their overall resilience.

Smaller, more rural and less wealthy states that lack the financial resources and logistical capabilities to respond effectively would be disproportionately affected.

“States don’t have that capability built to handle a disaster every single year,” Lynn Budd, director of the Wyoming Office of Homeland Security, told Stateline in an interview. Access to FEMA avoids the need for expensive disaster response infrastructure in each state.

States might be able to arrange regional cooperation. But state-led responses and regional models have limitations. The National Guard could assist with supply distribution, but it isn’t designed to provide fast financial aid, housing or long-term recovery options, and the supplies and the recovery effort still come at a cost.

Wealthier states might be better equipped to manage on their own, but poorer states would likely struggle. States with less funding and infrastructure would be left relying on nonprofits and community-based efforts. But these organizations are not capable of providing the scope of services FEMA can.

Any federal funding would also be slow if Congress had to approve aid after each disaster, rather than having FEMA already prepared to respond. States would be at the mercy of congressional infighting.

In the absence of a federal response and coordinating role, recovery would be uneven, with wealthier areas recovering faster and poorer areas likely seeing more prolonged hardships.

What does this mean?

Coordinating disaster response is complex, the paperwork for federal assistance can be frustrating, and the agency does draw criticism. However, it also fills an important role.

As the frequency of natural disasters continues to rise due to climate change, ask yourself: How prepared is your state for a disaster, and could it get by without federal aid?

Ming Xie, Assistant Professor of Emergency Management and Public Health, University of Maryland, Baltimore County

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Lu Wang, University of Michigan

AI has made it easier than ever to find information: Ask ChatGPT almost anything, and the system swiftly delivers an answer. But the large language models that power popular tools like OpenAI’s ChatGPT or Anthropic’s Claude were not designed to be accurate or factual. They regularly “hallucinate” and offer up falsehoods as if they were hard facts.

Yet people are relying more and more on AI to answer their questions. Half of all people in the U.S. between the ages of 14 and 22 now use AI to get information, according to a 2024 Harvard study. An analysis by The Washington Post found that more than 17% of prompts on ChatGPT are requests for information.

One way researchers are attempting to improve the information AI systems give is to have the systems indicate how confident they are in the accuracy of their answers. I’m a computer scientist who studies natural language processing and machine learning. My lab at the University of Michigan has developed a new way of deriving confidence scores that improves the accuracy of AI chatbot answers. But confidence scores can only do so much.

Popular and problematic

Leading technology companies are increasingly integrating AI into search engines. Google now offers AI Overviews that appear as text summaries above the usual list of links in any search result. Other upstart search engines, such as Perplexity, are challenging traditional search engines with their own AI-generated summaries.

The convenience of these summaries has made these tools very popular. Why scour the contents of multiple websites when AI can provide the most pertinent information in a few seconds?

AI tools seem to offer a smoother, more expedient avenue to getting information. But they can also lead people astray or even expose them to harmful falsehoods. My lab has found that even the most accurate AI models hallucinate in 25% of claims. This hallucination rate is concerning because other research suggests AI can influence what people think.

Language models hallucinate because they learn and operate on statistical patterns drawn from a massive amount of text data, much of which comes from the internet. This means that they are not necessarily grounded in real-world facts. They also lack other human competencies, like common sense and the ability to distinguish between serious expressions and sarcastic ones.

All this was on display last spring, when a user asked Google’s AI Overviews tool to suggest a way to keep cheese from sliding off a pizza. The tool promptly recommended mixing the cheese with glue. It then came to light that someone had once posted this obviously tongue-in-cheek recommendation on Reddit. Like most large language models, Google’s model had likely been trained with information scraped from myriad internet sources, including Reddit. It then mistakenly interpreted this user’s joke as a genuine suggestion.

While most users wouldn’t take the glue recommendation seriously, some hallucinated information can cause real harm. AI search engines and chatbots have repeatedly been caught citing debunked, racist pseudoscience as fact. Last year, Perplexity AI stated that a police officer in California was guilty of a crime that he did not commit.

Showing confidence

Building AI systems that prioritize veracity is challenging, but not impossible. One way AI developers are approaching this problem is to design models that communicate their confidence in their answers. This typically comes in the form of a confidence score – a number indicating how likely it is that a model is providing accurate information. But estimating a model’s confidence in the content it provides is also a complicated task.

One common approach to making this estimate involves asking the model to repeatedly respond to a given query. If the model is reliable, it should generate similar answers to the same query. If it can’t answer consistently, the AI is likely lacking the information it needs to answer accurately. Over time, the results of these tests become the AI’s confidence scores for specific subject areas.

Other approaches evaluate AI accuracy by directly prompting and training models to state how confident they are in their answers. But this offers no real accountability. Allowing an AI to evaluate its own confidence leaves room for the system to give itself a passing grade and continue to offer false or harmful information.

My lab has designed algorithms that assign confidence scores by breaking down a large language model’s responses into individual claims that can be automatically cross-referenced with Wikipedia. We assess the semantic equivalence between the AI model’s output and the referenced Wikipedia entries for the assertions. Our approach allows the AI to quickly evaluate the accuracy of all its statements. Of course, relying on Wikipedia articles, which are usually but not always accurate, also has its limitations.

Publishing confidence scores along with a model’s answers could help people to think more critically about the veracity of information that these tools provide. A language model can also be trained to withhold information if it earns a confidence score that falls below a set threshold. My lab has also shown that confidence scores can be used to help AI models generate more accurate answers.

Limits of confidence

There’s still a long way to go to ensure truly accurate AI. Most of these approaches assume that the information needed to correctly evaluate an AI’s accuracy can be found on Wikipedia and other online databases.

But when accurate information is just not that easy to come by, confidence estimates can be misleading. To account for cases like these, Google has developed special mechanisms for evaluating AI-generated statements. My lab has similarly compiled a benchmarking dataset of prompts that commonly cause hallucinations.

But all these approaches verify basic facts – there are no automated methods for evaluating other facets of long-form content, such as cause-and-effect relationships or an AI’s ability to reason over text consisting of more than one sentence.

Developing tools that improve these elements of AI are key steps toward making the technology a source of trustworthy information – and avoid the harms that misinformation can cause.

Lu Wang, Associate Professor of Computer Science and Engineering, University of Michigan

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Ryan Logan, UMass Chan Medical School and Mackenzie Gamble, Boston University

A good night’s sleep often sets the stage for a positive day. But for the nearly quarter of American adults struggling with mental illness, a good night’s rest is often elusive.

For patients with psychiatric conditions from addiction to mood disorders such as depression, disrupted sleep can often exacerbate symptoms and make it harder to stay on treatment.

Despite the important role circadian rhythms and sleep play in addiction, neuroscientists like me are only now beginning to understand the molecular mechanisms behind these effects.

Sleep and addictive drugs have an entangled relationship. Most addictive drugs can alter sleep-wake cycles, and sleep disorders in people using drugs are linked to addiction severity and relapse. While this poses a classic “chicken-or-egg” dilemma, it also presents an opportunity to understand how the sleep-addiction connection could unlock new treatments.

Circadian rhythms and health

At the center of the connection between sleep and mental health lies circadian rhythms: your body’s internal clock.

These rhythms align your bodily functions with your environment, synchronizing your body to day and night down to the molecular level. It does this through a series of proteins that interact in a feedback loop, turning genes on and off in regular patterns to support specific functions. Although your sleep-wake cycles are the most visible expression of circadian rhythms, these rhythms orchestrate most of your physiology.

If you have ever traveled across time zones, you have likely experienced a common form of circadian disruption called jet lag. This misalignment impairs your sleep and concentration, and can leave you feeling irritable.

While jet lag is a temporary nuisance, chronic circadian disruption such as frequent night shifts can lead to long-term health consequences, including an increased risk of cardiovascular disease and diabetes.

Circadian rhythms, sleep and opioid use

A major focus of my lab is on opioid addiction, a disease that has claimed nearly 80,000 lives a year since 2021 in the U.S. and has limited treatment options.

People addicted to opioids often experience disruptions to circadian rhythms, such as in their sleep and their levels of corticotropin, a key hormone that regulates stress. These disruptions are associated with many negative health consequences. In the short term, these disruptions can impair cognitive functions such as attention and increase negative emotions. Over time this can worsen mental and physical health. Studies of opioid addiction in mice reveal similar disruptions in sleep and various hormonal rhythms.

Importantly, poor sleep is common throughout a person’s experience with opioid use disorder, from actively using to withdrawal from opioids, and even while on treatment. This complication can have profound consequences. Studies have linked sleep disruption to a 2.5-fold increased risk of relapse among those undergoing treatment.

Unlocking the clock for opioid addiction

Using brain tissue from deceased donors and experiments in mice, my team is identifying molecular changes associated with psychiatric disorders in people. We model these changes in mice to explore how they affect disease severity and behavior.

Through genetic sequencing and computer modeling, my lab is able to profile all the RNA molecules in a brain region and understand how their rhythmicity – the peaks and troughs of their activity across the day – changes due to opioids. This provides a complete snapshot of which genes change at what time, allowing my team to peer into the molecular mechanics that may drive opioid addiction.

For example, we looked at two brain regions strongly associated with addiction: the nucleus accumbens and the dorsolateral prefrontal cortex. We found that patients with opioid addiction had completely different gene expression patterns in these brain regions compared with those without addiction. Some genes had adopted a completely different rhythm of activity, while others had lost their rhythmicity altogether.

Genes that lost rhythmicity included those involved in various components of the molecular clock and those linked to sleep duration. This further highlights how circadian disruption is a symptom of opioid use while beginning to uncover its underlying mechanisms.

In work that is pending peer review, my team focused on one major gene that lost rhythmicity in patients with opioid addiction: NPAS2. This component of the molecular clock is highly active in the nucleus accumbens and important for sleep and circadian regulation. We found that blocking functional NPAS2 formation led to increased fentanyl-seeking behavior in mice. Interestingly, we observed that female mice were willing to press a lever more times than male mice to obtain fentanyl, reflecting documented sex differences in opioid addiction among people. In another study, we also found that lack of NPAS2 exacerbated sleep disruption in mice that were administered fentanyl.

Together, our findings reinforce the role circadian rhythms play in addiction. Future work may clarify whether targeting NPAS2 could treat opioid addiction symptoms. Quality sleep isn’t just about waking up refreshed – it could also lead to reduced opioid use and fewer overdoses.

Ryan Logan, Professor of Psychiatry, UMass Chan Medical School and Mackenzie Gamble, Ph.D. Candidate in Molecular and Translational Medicine, Boston University

This article is republished from The Conversation under a Creative Commons license. Read the original article.