Christina Erickson, University of North Dakota

Nearly a half-century after the Supreme Court ruled that school spankings are permissible and not “cruel and unusual punishment”, many U.S. states allow physical punishment for students who have misbehaved.

Today, over a third of the states allow teachers to paddle or spank students. More than 100,000 students are paddled in U.S. schools each year.

Christina Erickson, an associate dean and professor of social work at the University of North Dakota, wrote a book on the subject: “Spanked: How Hitting Our Children is Harming Ourselves.” She discussed the scope of the practice and its effects with The Conversation.

What spanking legislation exists worldwide?

Around the world, 68 countries have banned the hitting of children in any form, including spanking. This movement began in 1979 with Sweden’s ban on all forms of physical punishment, including spanking in any setting, and including in the family home.

The pace of change quickened in the early 2000s when more countries adopted similar laws. For example, the legal language of countries like Nepal rests on an emerging definition of children as rights holders similar to adults and as humans worth protecting from harm.

What are US policies toward spanking?

Each state in the U.S. has its own child abuse laws, and all states, tribes and territories aim to protect children from abuse. But all state laws also allow parents to hit their children if it does not leave an injury or a mark.

A typical example is Oklahoma’s definition of child abuse and neglect. It includes an exception that permits parents to use ordinary force as a means of discipline, including spanking, using an implement like a switch or a paddle. However, leaving evidence of hitting, such as welts, bruises, swelling or lacerations, is illegal and considered child abuse in all states.

Parental spanking of children is considered unique from other physical violence because of the relational context and the purpose. Laws entitle parents to hit their children for the purpose of teaching a lesson or punishing them to improve behavior. Children are the only individuals in society who can be hit by another person and the law does not regard it as assault.

Spanking’s impact on a child is unfortunately similar to abusive hitting. Spanking has been labeled as an “Adverse Childhood Experience,” or ACE. These are events that cause poor health outcomes over the span of one’s life.

The practice of spanking also affects parents. Acceptance of the physical discipline of spanking puts parents at risk for the escalation of physical punishment that leads to abuse.

Parents who spank their child have the potential to abuse them and be caught in a legal and child protection system that aims to protect children from harm. It is unclear what triggers a parent to cross over from discipline into abuse. Research shows that spanking at a young age, such as a 1-year-old, increases the chance of involvement by Child Protective Services by 33%.

Some school districts require permission from parents to allow disciplinary paddling in school, while others do not require any communication. State law does not assure agreement between parents and school districts on what offenses warrant a paddling. Parents may feel they have no alternative but to keep their child in school, or fear reprisal from school administrators. Some students are old enough to denounce the punishment themselves.

Is spanking considered the same as hitting?

The term spank conceals the concept of hitting and is so commonplace it goes unquestioned, despite the fact that it is a grown adult hitting a person much smaller than them. The concept is further concealed because hitting a child’s bottom hides any injuries that may occur.

Types of hitting that are categorized as spanking have narrowed over the years but still persist. Some parents still use implements such as tree switches, wooden spoons, shoes or paddles to “spank” children, raising the chances for abuse.

Most spanking ends by the age of 12, partly because children this age are able to fight back. When a child turns 18, parental hitting becomes the same as hitting any other adult, a form of domestic violence or assault throughout the U.S.

There is a lack of a consistent understanding of what constitutes a spanking. The definition of spanking is unique to each family. The number of hits, clothed or not, or using an implement, all reflect geographical or familial differences in understanding what a spanking is.

How do US adults view spanking?

People in the United States generally accept spanking as part of raising children: 56% of U.S. adults strongly agree or agree that “… it is sometimes necessary to discipline a child with a good, hard spanking.” This view has been slowly changing since 1986, when 83% of adults agreed with that statement.

The laws worldwide that protect children from being hit usually begin by disallowing nonparental adults to hit children. This is happening in the U.S. too, where 31 states have banned paddling in schools.

At a national level, efforts have been made to end physical punishment in schools. However, 19 states still allow spanking of children in public schools, which was upheld by a 1977 Supreme Court case.

With the slow but steady drop of parents who believe that sometimes children need a good hard spanking, as well as the ban of paddling in schools in 31 states, one could argue that the U.S. is moving toward a reduction in spanking.

What does research say about spanking?

Spanking’s negative influence on children’s behavior has been documented for decades. Spanking seems to work in the moment when it comes to changing or stopping the immediate behavior, but the negative effects are hidden in the short term and occur later in the child’s life. Yet because the spanking seemed to work at the time, the parent doesn’t connect the continued bad behavior of the child to the spanking.

An abundance of research shows that spanking causes increased negative behaviors in childhood. Spanking lowers executive functioning for children, increases dating violence as teenagers and even increases struggles with mental health and substance abuse in adulthood. Spanking does not teach new or healthy behaviors, and is a stress-inducing event for the child and the adult hitting them.

No studies have shown positive long-term benefits from spanking. Because of the long-standing and expansive research findings showing a range of harm from spanking and the increased association with child abuse, the American Psychological Association recommends that parents should never spank their children.

What are some resources for parents?

Consider these questions when choosing a discipline method for your child:

• Is the expectation of your child developmentally accurate? One of the most common reasons parents spank is because they are expecting a behavior the child is not developmentally able to execute.

• Can the discipline you choose grow with your child? Nearly all spanking ends by age 12, when kids are big enough to fight back. Choose discipline methods you can use over the long term, such as additional chores, apologies, difficult conversations and others that can grow with your child.

• Might there be another explanation for your child’s behavior? Difficulty of understanding, fear or miscommunication? Think of your child as a learner and use a growth mindset to help your child learn from their life experiences.

Parents are the leaders of their families. Good leaders show strength in nonthreatening ways, listen to others and explain their decisions. Don’t spoil your kids. But being firm does not have to include hitting.

Is spanking children good for parents?

Doubtful. Parents who hit their kids may be unaware that it influences their frustration in other relationships. Expressing aggression recharges an angry and short-tempered internal battery that transfers into other parts of the adults’ lives.

Practicing calm when with your children will help you be calmer at work and in your other relationships. Listening to and speaking with a child about challenges, even from a very early age, is the best way to make it part of your relationship for the rest of your life.

Choose a method that allows you to grow. Parents matter too.

Christina Erickson, Associate Dean in the College of Nursing and Professional Disciplines, University of North Dakota

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

Manas Gaur, University of Maryland, Baltimore County

ChatGPT and other AI chatbots based on large language models are known to occasionally make things up, including scientific and legal citations. It turns out that measuring how accurate an AI model’s citations are is a good way of assessing the model’s reasoning abilities.

An AI model “reasons” by breaking down a query into steps and working through them in order. Think of how you learned to solve math word problems in school.

Ideally, to generate citations an AI model would understand the key concepts in a document, generate a ranked list of relevant papers to cite, and provide convincing reasoning for how each suggested paper supports the corresponding text. It would highlight specific connections between the text and the cited research, clarifying why each source matters.

The question is, can today’s models be trusted to make these connections and provide clear reasoning that justifies their source choices? The answer goes beyond citation accuracy to address how useful and accurate large language models are for any information retrieval purpose.

I’m a computer scientist. My colleagues − researchers from the AI Institute at the University of South Carolina, Ohio State University and University of Maryland Baltimore County − and I have developed the Reasons benchmark to test how well large language models can automatically generate research citations and provide understandable reasoning.

We used the benchmark to compare the performance of two popular AI reasoning models, DeepSeek’s R1 and OpenAI’s o1. Though DeepSeek made headlines with its stunning efficiency and cost-effectiveness, the Chinese upstart has a way to go to match OpenAI’s reasoning performance.

Sentence specific

The accuracy of citations has a lot to do with whether the AI model is reasoning about information at the sentence level rather than paragraph or document level. Paragraph-level and document-level citations can be thought of as throwing a large chunk of information into a large language model and asking it to provide many citations.

In this process, the large language model overgeneralizes and misinterprets individual sentences. The user ends up with citations that explain the whole paragraph or document, not the relatively fine-grained information in the sentence.

Further, reasoning suffers when you ask the large language model to read through an entire document. These models mostly rely on memorizing patterns that they typically are better at finding at the beginning and end of longer texts than in the middle. This makes it difficult for them to fully understand all the important information throughout a long document.

Large language models get confused because paragraphs and documents hold a lot of information, which affects citation generation and the reasoning process. Consequently, reasoning from large language models over paragraphs and documents becomes more like summarizing or paraphrasing.

The Reasons benchmark addresses this weakness by examining large language models’ citation generation and reasoning.

Testing citations and reasoning

Following the release of DeepSeek R1 in January 2025, we wanted to examine its accuracy in generating citations and its quality of reasoning and compare it with OpenAI’s o1 model. We created a paragraph that had sentences from different sources, gave the models individual sentences from this paragraph, and asked for citations and reasoning.

To start our test, we developed a small test bed of about 4,100 research articles around four key topics that are related to human brains and computer science: neurons and cognition, human-computer interaction, databases and artificial intelligence. We evaluated the models using two measures: F-1 score, which measures how accurate the provided citation is, and hallucination rate, which measures how sound the model’s reasoning is − that is, how often it produces an inaccurate or misleading response.

Our testing revealed significant performance differences between OpenAI o1 and DeepSeek R1 across different scientific domains. OpenAI’s o1 did well connecting information between different subjects, such as understanding how research on neurons and cognition connects to human-computer interaction and then to concepts in artificial intelligence, while remaining accurate. Its performance metrics consistently outpaced DeepSeek R1’s across all evaluation categories, especially in reducing hallucinations and successfully completing assigned tasks.

OpenAI o1 was better at combining ideas semantically, whereas R1 focused on making sure it generated a response for every attribution task, which in turn increased hallucination during reasoning. OpenAI o1 had a hallucination rate of approximately 35% compared with DeepSeek R1’s rate of nearly 85% in the attribution-based reasoning task.

In terms of accuracy and linguistic competence, OpenAI o1 scored about 0.65 on the F-1 test, which means it was right about 65% of the time when answering questions. It also scored about 0.70 on the BLEU test, which measures how well a language model writes in natural language. These are pretty good scores.

DeepSeek R1 scored lower, with about 0.35 on the F-1 test, meaning it was right about 35% of the time. However, its BLEU score was only about 0.2, which means its writing wasn’t as natural-sounding as OpenAI’s o1. This shows that o1 was better at presenting that information in clear, natural language.

OpenAI holds the advantage

On other benchmarks, DeepSeek R1 performs on par with OpenAI o1 on math, coding and scientific reasoning tasks. But the substantial difference on our benchmark suggests that o1 provides more reliable information, while R1 struggles with factual consistency.

Though we included other models in our comprehensive testing, the performance gap between o1 and R1 specifically highlights the current competitive landscape in AI development, with OpenAI’s offering maintaining a significant advantage in reasoning and knowledge integration capabilities.

These results suggest that OpenAI still has a leg up when it comes to source attribution and reasoning, possibly due to the nature and volume of the data it was trained on. The company recently announced its deep research tool, which can create reports with citations, ask follow-up questions and provide reasoning for the generated response.

The jury is still out on the tool’s value for researchers, but the caveat remains for everyone: Double-check all citations an AI gives you.

Manas Gaur, Assistant Professor of Computer Science and Electrical Engineering, University of Maryland, Baltimore County

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

Hanna D. Paton, University of Iowa

The flu sickens millions of people in the U.S. every year, and the past year has been particularly tough. Although infections are trending downward, the Centers for Disease Control and Prevention has called the winter of 2024-2025 a “high severity” season with the highest hospitalization rate in 15 years.

Since early 2024, a different kind of flu called bird flu, formally known as avian influenza, has been spreading in birds as well as in cattle. The current bird flu outbreak has infected 70 Americans and caused two deaths as of April 8, 2025. Public health and infectious disease experts say the risk to people is currently low, but they have expressed concern that this strain of the bird flu virus may mutate to spread between people.

As a doctoral candidate in immunology, I study how pathogens that make us sick interact with our immune system. The viruses that cause seasonal flu and bird flu are distinct but still closely related. Understanding their similarities and differences can help people protect themselves and their loved ones.

What is influenza?

The flu has long been a threat to public health. The first recorded influenza pandemic occurred in 1518, but references to illnesses possibly caused by influenza stretch back as as early as 412 B.C., to a treatise called Of the Epidemics by the Greek physician Hippocrates.

Today, the World Health Organization estimates that the flu infects 1 billion people every year. Of these, 3 million to 5 million infections cause severe illness, and hundreds of thousands are fatal.

Influenza is part of a large family of viruses called orthomyxoviruses. This family contains several subtypes of influenza, referred to as A, B, C and D, which differ in their genetic makeup and in the types of infections they cause. Influenza A and B pose the largest threat to humans and can cause severe disease. Influenza C causes mild disease, and influenza D is not known to infect people. Since the turn of the 20th century, influenza A has caused four pandemics. Influenza B has never caused a pandemic.

An influenza A strain called H1N1 caused the famous 1918 Spanish flu pandemic, which killed about 50 million people worldwide. A related H1N1 virus was responsible for the most recent influenza A pandemic in 2009, commonly referred to as the swine flu pandemic. In that case, scientists believe multiple different types of influenza A virus mixed their genetic information to produce a new and especially virulent strain of the virus that infected more than 60 million people in the U.S. from April 12, 2009, to April 10, 2010, and caused huge losses to the agriculture and travel industries.

Both swine and avian influenza are strains of influenza A. Just as swine flu strains tend to infect pigs, avian flu strains tend to infect birds. But the potential for influenza A viruses that typically infect animals to cause pandemics in humans like the swine flu pandemic is why experts are concerned about the current avian influenza outbreak.

Seasonal flu versus bird flu

Different strains of influenza A and influenza B emerge each year from about October to May as seasonal flu. The CDC collects and analyzes data from public health and clinical labs to determine which strains are circulating through the population and in what proportions. For example, recent data shows that H1N1 and H3N2, both influenza A viruses, were responsible for the vast majority of cases this season. Standard tests for influenza generally determine whether illness is caused by an A or B strain, but not which strain specifically.

Officials at the Food and Drug Administration use this information to make strain recommendations for the following season’s influenza vaccine. Although the meeting at which FDA advisers were to decide the makeup of the 2026 flu vaccine was unexpectedly canceled in late February, the FDA still released its strain recommendations to manufacturers.

The recommendations do not include H5N1, the influenza A strain that causes avian flu. The number of strains that can be added into seasonal influenza vaccines is limited. Because cases of people infected with H5N1 are minimal, population-level vaccination is not currently necessary. As such, seasonal flu vaccines are not designed to protect against avian influenza. No commercially available human vaccines currently exist for avian influenza viruses.

How do people get bird flu?

Although H5N1 mainly infects birds, it occasionally infects people, too. Human cases, first reported in 1997 in Hong Kong, have primarily occurred in poultry farm workers or others who have interacted closely with infected birds.

Initially identified in China in 1996, the first major outbreak of H5 family avian flu occurred in North America in 2014-2015. This 2014 outbreak was caused by the H5N8 strain, a close relative of H5N1. The first H5N1 outbreak in North America began in 2021 when infected birds carried the virus across the ocean. It then ripped through poultry farms across the continent.

In March 2024, epidemiologists identified H5N1 infections in cows on dairy farms. This is the first time that bird flu was reported to infect cows. Then, on April 1, 2024, health officials in Texas reported the first case of a person catching bird flu from infected cattle. This was the first time transmission of bird flu between mammals was documented.

As of March 21, 2025, there have been 988 human cases of H5N1 worldwide since 1997, about half of which resulted in death. The current outbreak in the U.S. accounts for 70 of those infections and one death. Importantly, there have been no reports of H5N1 spreading directly from one person to another.

Since avian flu is an influenza A strain, it would show up as positive on a standard rapid flu test. However, there is no evidence so far that avian flu is significantly contributing to current influenza cases. Specific testing is required to confirm that a person has avian flu. This testing is not done unless there is reason to believe the person was exposed to sick birds or other sources of infection.

How might avian flu become more dangerous?

As viruses replicate within the cells of their host, their genetic information can get copied incorrectly. Some of these genetic mutations cause no immediate differences, while others alter some key viral characteristics.

Influenza viruses mutate in a special way called reassortment, which occurs when multiple strains infect the same cell and trade pieces of their genome with one another, potentially creating new, unique strains. This process prolongs the time the virus can inhabit a host before an infection is cleared. Even a slight change in a strain of influenza can result in the immune system’s inability to recognize the virus. As a result, this process forces our immune systems to build new defenses instead of using immunity from previous infections.

Reassortment can also change how harmful strains are to their host and can even enable a strain to infect a different species of host. For example, strains that typically infect pigs or birds may acquire the ability to infect people. Influenza A can infect many different types of animals, including cattle, birds, pigs and horses. This means there are many strains that can intermingle to create novel strains that people’s immune systems have not encountered before – and are therefore not primed to fight.

It is possible for this type of transformation to also occur in H5N1. The CDC monitors which strains of flu are circulating in order prepare for that possibility. Additionally, the U.S. Department of Agriculture has a surveillance system for monitoring potential threats for spillover from birds and other animals, although this capacity may be at risk due to staff cuts in the department.

These systems are critical to ensure that public health officials have the most up-to-date information on the threat that H5N1 poses to public health and can take action as early as possible when a threat is evident.

Hanna D. Paton, PhD Candidate in Immunology, University of Iowa

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