Tuesday, December 2, 2025

Early Smartphone Use Associated With Sleep Problems and Mental Health Issues in Preteens

Twelve-year-olds who owned smartphones had higher odds of depression, obesity, and sleep problems compared to peers without smartphones, according to a study published December 1, 2025, in the journal Pediatrics.

Researchers analyzed data from 10,588 children and adolescents across 21 sites between 2016 and 2022. The participants were part of the Adolescent Brain Cognitive Development Study, funded by the National Institutes of Health. The research team included scientists from Children's Hospital of Philadelphia, the University of Pennsylvania, the University of California at Berkeley, and Columbia University.

At age 12, children who owned smartphones had 30 percent higher odds of depression, 40 percent higher odds of obesity, and 60 percent higher odds of insufficient sleep compared to those without smartphones. The study defined insufficient sleep as less than nine hours per day.

The data showed 64 percent of participants owned smartphones at age 12. The median age for receiving a first smartphone was 11 years. At age 14, smartphone ownership had reached 89 percent.

Researchers found that the age when children first received smartphones mattered. For each year younger a child was when receiving their first smartphone, the risk of obesity and insufficient sleep at age 12 increased by approximately 8 to 9 percent. This pattern held even for children as young as age 4.

The study included a separate analysis of 3,486 children who did not own smartphones at age 12. Among these children, 1,546 acquired smartphones within the following year while 1,940 did not. At age 13, those who had acquired smartphones had 57 percent higher odds of clinical-level mental health problems and 50 percent higher odds of insufficient sleep compared to those who remained without smartphones. These results accounted for baseline mental health and sleep measures at age 12.

The researchers controlled for multiple factors including age, sex, income, parental education, race, and ethnicity. They also adjusted for ownership of other devices such as tablets, pubertal development, and parental monitoring. Results remained consistent across several different analytical approaches.

Dr. Ran Barzilay, the study's lead author and a child and adolescent psychiatrist at Children's Hospital of Philadelphia, noted the research examined only whether owning a smartphone was associated with health outcomes. The study did not investigate what children were "doing on their smartphones".

The researchers accounted for children's use of other technological devices including tablets and iPads. These adjustments did not change the findings.

The study could not determine whether smartphones directly caused these health problems. Previous research has found that excessive smartphone use correlates with reduced in-person social interactions, less physical activity, and decreased sleep, all of which can affect adolescent health.

Barzilay stated the findings showed health impacts even when smartphone use was not considered problematic. He emphasized that smartphones can serve beneficial purposes by strengthening social connections and supporting learning. Some families consider smartphones necessary for their children's safety.

Children between ages 8 and 12 average slightly over five hours of screen time per day, according to data cited in the study.

The researchers called for additional studies to identify which specific aspects of smartphone ownership and use connect to negative health outcomes. They plan to examine younger children who received smartphones before age 10 to understand who faces the greatest vulnerability to harmful effects and who might benefit most from smartphone access.
The study authors recommended that parents, children, and pediatricians engage in careful discussions before children receive smartphones. Barzilay suggested parents can implement rules such as prohibiting phone use in bedrooms at night and ensuring children participate in activities that do not require phones. He advised parents to monitor phone content and prevent smartphones from disrupting sleep.

The researchers noted their findings should inform both family decisions about smartphone use and potential public policy aimed at protecting youth health. They emphasized that some children who do not own smartphones may face various adverse consequences and challenges, highlighting the need to support families navigating this decision.


Notes: This post was drafted with the assistance of AI tools and reviewed, edited, and published by humans. Image: DIW-Aigen.

Read next: Global Smartphone Market to Grow in 2025 as Memory Shortage Drives Price Pressures for 2026
by Asim BN via Digital Information World

Global Smartphone Market to Grow in 2025 as Memory Shortage Drives Price Pressures for 2026

Worldwide smartphone shipments are expected to rise in 2025, with IDC forecasting a 1.5% year-over-year increase to about 1.25 billion units. The improved outlook reflects stronger demand for Apple devices, firmer results in key emerging markets, and steadier conditions in China. IDC raised its forecast following Apple’s faster-than-expected momentum entering the holiday quarter.

Apple’s performance accounts for a substantial part of the improved forecast. IDC expects the company to ship 247.4 million iPhones next year, reflecting 6.1% annual growth and marking its highest volume on record. China contributes significantly to this shift. IDC revised Apple’s 2025 outlook for the region from a projected 1% decline to 3% growth after recent monthly sales data showed sustained demand. Globally, Apple’s shipment value is projected to exceed 261 billion dollars in 2025, supported by 7.2% year-over-year growth.


The outlook changes in 2026 as component availability tightens. IDC now expects a 0.9% decline in worldwide smartphone shipments, reversing an earlier projection for slight growth. The revision reflects two factors: a global memory shortage that is raising costs and constraining supply, and Apple’s decision to move the launch of its next base model from late 2026 to early 2027. IDC notes that the shortage is expected to affect lower-end and midrange Android devices more noticeably because they are more sensitive to price increases.

Pricing is expected to rise even as unit volumes soften. IDC forecasts the global average selling price of smartphones to reach 465 dollars in 2026. Higher component costs are expected to push overall market value to 578.9 billion dollars. Manufacturers may raise retail prices or adjust their portfolios toward higher-margin models to manage the impact of memory-related cost increases.

The market enters 2025 with improving conditions, while the balance between component constraints and pricing trends shapes expectations for 2026.

Notes: This post was drafted with the assistance of AI tools and reviewed, edited, and published by humans.

Read next:

• How Small Language Models Differ from Large Ones in Power and Purpose

• Microsoft CEO on the Skills That Matter as AI Expands in the Workplace
by Irfan Ahmad via Digital Information World

How Small Language Models Differ from Large Ones in Power and Purpose

Microsoft just released its latest small language model that can operate directly on the user’s computer. If you haven’t followed the AI industry closely, you might be asking: what exactly is a small language model (SLM)?

As AI becomes increasingly central to how we work, learn and solve problems, understanding the different types of AI models has never been more important. Large language models (LLMs) such as ChatGPT, Claude, Gemini and others are in widespread use. But small ones are increasingly important, too.

Image: DIW-Aigen

Let’s explore what makes SLMs and LLMs different – and how to choose the right one for your situation.

Firstly, what is a language model?

You can think of language models as incredibly sophisticated pattern-recognition systems that have learned from vast amounts of text.

They can understand questions, generate responses, translate languages, write content, and perform countless other language-related tasks.

The key difference between small and large models lies in their scope, capability and resource requirements.

Small language models are like specialised tools in a toolbox, each designed to do specific jobs extremely well. They typically contain millions to tens of millions of parameters (these are the model’s learned knowledge points).

Large language models, on the other hand, are like having an entire workshop at your disposal – versatile and capable of handling almost any challenge you throw at them, with billions or even trillions of parameters.

What can LLMs do?

Large language models represent the current pinnacle of AI language capabilities. These are the models making headlines for their ability to “write” poetry, debug complex code, engage in conversation, and even help with scientific research.

When you interact with advanced AI assistants such as ChatGPT, Gemini, Copilot or Claude, you’re experiencing the power of LLMs.

The primary strength of LLMs is their versatility. They can handle open-ended conversations, switching seamlessly from discussing marketing strategies to explaining scientific concepts to creative writing. This makes them invaluable for businesses that need AI to handle diverse, unpredictable tasks.

A consulting firm, for instance, might use an LLM to analyse market trends, generate comprehensive reports, translate technical documents, and assist with strategic planning – all with the same model.

LLMs excel at tasks requiring nuanced understanding and complex reasoning. They can interpret context and subtle implications, and generate responses that consider multiple factors simultaneously.

If you need AI to review legal contracts, synthesise information from multiple sources, or engage in creative problem-solving, you need the sophisticated capabilities of an LLM.

These models are also excellent at generalising. Train them on diverse data, and they can extrapolate knowledge to handle scenarios they’ve never explicitly encountered.

However, LLMs require significant computational power and usually run in the cloud, rather than on your own device or computer. In turn, this translates to high operational costs. If you’re processing thousands of requests daily, these costs can add up quickly.

When less is more: SLMs

In contrast to LLMs, small language models excel at specific tasks. They’re fast, efficient and affordable.

Take a library’s book recommendation system. An SLM can learn the library’s catalogue. It “understands” genres, authors and reading levels so it can make great recommendations. Because it’s so small, it doesn’t need expensive computers to run.

SLMs are easy to fine-tune. A language learning app can teach an SLM about common grammar mistakes. A medical clinic can train one to understand appointment scheduling. The model becomes an expert in exactly what you need.

SLMs are faster than LLMs, too – they can deliver answers in milliseconds, rather than seconds. This difference may seem small, but it’s noticeable in applications such as grammar checkers or translation apps, which can’t keep users waiting.

Costs are much smaller, too. Small language models are like LED bulbs – efficient and affordable. Large language models are like stadium lights – powerful but expensive.

Schools, non-profits and small businesses can use SLMs for specific tasks without breaking the bank. For example, Microsoft’s Phi-3 small language models are helping power an agricultural information platform in India to provide services to farmers even in remote places with limited internet.

SLMs are also great for constrained systems such as self-driving cars or satellites that have limited processing power, minimal energy budgets, and no reliable cloud connection. LLMs simply can’t run in these environments. But an SLM, with its smaller footprint, can fit onboard.

Both types of models have their place

What’s better – a minivan or a sports car? A downtown studio apartment or a large house in the suburbs? The answer, of course, is that it depends on your needs and your resources.

The landscape of AI models is rapidly evolving, and the line between small and large models is becoming increasingly nuanced. We’re seeing hybrid approaches where businesses use SLMs for routine tasks and escalate to LLMs for complex queries. This approach optimises both cost and performance.

The choice between small and large language models isn’t about which is objectively better – it’s about which better serves your specific needs.

SLMs offer efficiency, speed and cost-effectiveness for focused applications, making them ideal for businesses with specific use cases and resource constraints.

LLMs provide unmatched versatility and sophistication for complex, varied tasks, justifying their higher resource requirements when a highly capable AI is needed.

Lin Tian, Research Fellow, Data Science Institute, University of Technology Sydney and Marian-Andrei Rizoiu, Associate Professor in Behavioral Data Science, University of Technology Sydney

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

Read next: “Rage Bait” Named Oxford Word of the Year 2025


by Web Desk via Digital Information World

Microsoft CEO on the Skills That Matter as AI Expands in the Workplace

As artificial intelligence increasingly handles technical tasks, Microsoft CEO Satya Nadella emphasized that human skills remain essential for workplace success. In a conversation with Axel Springer CEO Mathias Döpfner that aired November 29, 2024 on the MD Meets podcast, Nadella highlighted the growing importance of emotional intelligence and collaboration.


Nadella noted that cognitive ability alone is insufficient for leaders and employees. He stated that emotional intelligence and social awareness are becoming more critical as AI automates routine responsibilities. Nadella explained that possessing intellectual capability without emotional intelligence diminishes its value. The workplace is increasingly a space where human interaction and collective problem-solving define outcomes.

When Döpfner asked whether empathy considerations were driving Microsoft to call more people back to the office, Nadella acknowledged the value of physical workspaces for collaboration but emphasized flexibility. While something important gets lost when people don't come together in person, Microsoft maintains a balanced approach rather than imposing rigid mandates. Physical spaces remain valuable for picking up social and emotional cues that enable better innovation and allow humans to accumulate knowledge through context that AI systems have not yet learned.

When asked whether companies could be entirely run by AI, Nadella described the notion as too far-fetched to imagine. He emphasized that human judgment, empathy, and decision-making remain irreplaceable. While AI can augment productivity, leadership and collaborative problem-solving cannot be fully replicated by machines. Nadella described a future work model involving macro delegation to AI agents that handle tasks but return for human guidance and micro steering when they encounter limitations or need direction.

Nadella stressed that successful AI implementation requires four elements. Organizations need a mindset embracing business process re-engineering rather than simply applying AI to existing workflows. They need appropriate tools, the skills to apply those tools effectively, and properly normalized data sets spanning multiple systems. Without this combination, AI projects will likely fail, which may explain why many executives expect productivity gains from AI but few have realized them.

Nadella's remarks reflect a broader perspective on AI adoption. Technology can enhance human capabilities, but leadership and empathy remain central to workplace effectiveness. Even in highly automated environments, human collaboration and understanding continue to shape business outcomes.

Notes: This post was drafted with the assistance of AI tools and reviewed, edited, and published by humans.

Read next: Threads Code Reveals AI Tool Designed to Summarize Profile Engagement Patterns
by Ayaz Khan via Digital Information World

Threads Code Reveals AI Tool Designed to Summarize Profile Engagement Patterns

Meta’s Threads app appears to be developing an AI tool that generates summaries of user interactions on profile pages, according to app researcher Alessandro Paluzzi.

The feature provides visitors with a snapshot of past engagements, including related interests or general activity patterns. It offers a quick overview without scrolling through individual posts or replies, resembling communication summaries on other platforms.
Potential Threads AI tool may summarize user engagement patterns, offering quick context despite announcements regarding functionality.
Paluzzi’s findings suggest these summaries could appear for any profile, even without prior interactions. The full functionality remains unconfirmed, and Threads has made no official announcement.

Threads’ development occurs alongside profile transparency tools on other platforms. X recently introduced a feature showing account details such as location, join date and username changes. The tool aims to reduce inauthentic engagement and is not AI-powered.

Commentators under Paluzzi post have noted potential implications of Threads’ summaries, including influencing engagement decisions or highlighting repeated critical interactions. These observations reflect user commentary rather than confirmed outcomes.

No official purposes or verified results are available beyond code analysis and internal testing reports. How the feature would function if released or whether it would be widely deployed remains unknown.

If implemented, the AI tool would allow visitors to quickly understand prior engagement patterns without manually reviewing past activity.
Observers note that while the summaries could streamline discovery, they might also subtly shape perceptions of profiles. Any impact on following or engagement is speculative and unconfirmed by Meta.

Threads’ exploration of AI-assisted summaries reflects a broader trend in social media toward tools that provide context and simplify interaction history. The feature remains experimental, with release timing and full functionality still unknown.

Notes: This post was drafted with the assistance of AI tools and reviewed, edited, and published by humans.

Read next: “Rage Bait” Named Oxford Word of the Year 2025
by Asim BN via Digital Information World

“Rage Bait” Named Oxford Word of the Year 2025

Oxford University Press has selected “rage bait” as its Word of the Year for 2025. The term refers to online content deliberately designed to provoke anger or outrage, typically posted to increase traffic or engagement on a website or social media account.

The phrase combines “rage,” meaning a violent outburst of anger, and “bait,” an attractive morsel of food. Although technically two words, Oxford lexicographers treat it as a single unit of meaning, showing how English adapts existing words to express new ideas.

The first recorded use of “rage bait” was in 2002 on Usenet, describing a driver’s reaction to being flashed by another driver. Over time, it evolved into internet slang for content intended to elicit anger, including viral social media posts.

Usage of the term has tripled in the past 12 months, indicating its growing presence in online discourse. Experts note that the word reflects how people interact with and respond to online content.

The Word of the Year was chosen through a combination of public voting and expert review. Two other words were shortlisted: “aura farming,” defined as cultivating an attractive or charismatic persona, and “biohack,” describing efforts to optimize physical or mental performance, health, or wellbeing through lifestyle, diet, supplements, or technology.

Casper Grathwohl, President of Oxford Languages, said the increase in usage highlights growing awareness of the ways online content can influence attention and behavior. He also compared “rage bait” to last year’s Word of the Year, “brain rot,” which described the mental drain of endless scrolling.

The annual Word of the Year reflects terms that captured significant cultural and linguistic trends over the previous 12 months, based on usage data, public engagement, and expert analysis.


Notes: This post was drafted with the assistance of AI tools and reviewed, edited, and published by humans. Image: DIW-Aigen.

Read next: Which AI Models Answer Most Accurately, and Which Hallucinate Most? New Data Shows Clear Gaps
by Irfan Ahmad via Digital Information World

Monday, December 1, 2025

Which AI Models Answer Most Accurately, and Which Hallucinate Most? New Data Shows Clear Gaps

Recent findings from the European Broadcasting Union show that AI assistants misrepresent news content in 45% of the test cases, regardless of language or region. That result underscores why model accuracy and reliability remain central concerns. Fresh rankings from Artificial Analysis, based on real-world endpoint testing as of 1 December 2025, give a clear picture of how today’s leading systems perform when answering direct questions.

Measuring Accuracy and Hallucination Rates

Artificial Analysis evaluates both proprietary and open weights models through live API endpoints. Their measurements reflect what users experience in actual deployments rather than theoretical performance. Accuracy shows how often a model produces correct answers. Hallucination rate captures how often it responds incorrectly when it should refuse or indicate uncertainty. Since new models launch frequently and providers adjust endpoints, these results can change over time, but the current snapshot still reveals clear trends.

Models With the Highest Hallucination Rates

Hallucination Metrics Expose Deep Reliability Risks in Current AI Assistant Deployments
Model Hallucination Rate
Claude 4.5 Haiku 26%
Claude 4.5 Sonnet 48%
GPT-5.1 (high) 51%
Claude Opus 4.5 58%
Magistral Medium 1.2 60%
Grok 4 64%
Kimi K2 0905 69%
Grok 4.1 Fast 72%
Kimi K2 Thinking 74%
Llama Nemotron Super 49B v1.5 76%
DeepSeek V3.2 Ex 81%
DeepSeek R1 0528 83%
EXAONE 4.032B 86%
Llama 4 Maverick 87.58%
Gemini 3 Pro Preview (high) 87.99%
Gemini 2.5 Flash (Sep) 88.31%
Gemini 2.5 Pro 88.57%
MiniMax-M2 88.88%
GPT-5.1 89.17%
Qwen3 235B A22B 2507 89.64%
gpt-oss-120B (high) 89.96%
GLM-4.6 93.09%
gpt-oss-20B (high) 93.20%

When it comes to hallucination, the gap between models is striking. Claude 4.5 Haiku has the lowest hallucination rate in this group at 26 percent, yet even this relatively low figure indicates that incorrect answers are common. Several models climb sharply from there. Claude 4.5 Sonnet reaches 48 percent, GPT-5.1 (High) 51 percent, and Claude Opus 4.5 58 percent. Grok 4 produces incorrect responses 64 percent of the time, and Kimi K2 0905 rises to 69 percent. Beyond these, models enter the seventies and eighties. Grok 4.1 Fast shows a 72 percent rate, Kimi K2 Thinking 74 percent, and Llama Nemotron Super 49B v1.5 76 percent. DeepSeek benchmarks show even higher rates, with V3.2 Ex at 81 percent and R1 0528 at 83 percent. Among the highest are EXAONE 4.032B at 86 percent, Llama 4 Maverick at 87.58 percent, and several Gemini models including 3 Pro Preview (High) and 2.5 Flash (Sep) exceeding 87 percent. GLM-4.6 and gpt-oss-20B (High) top the chart at over 93 percent. This spread demonstrates that while some models are relatively restrained, many generate incorrect answers frequently, making hallucination a major challenge for AI systems today.

Top Performers in Accuracy

Testing Reveals Limited Accuracy Gains Despite Rapid Deployment of Advanced AI Systems
Model Accuracy
Gemini 3 Preview (High) 54%
Claude Opus 4.5 43%
Grok 4 40%
Gemini 2.5 Pro 37%
GPT-5.1 (High) 35%
Claude 4.5 Sonnet 31%
DeepSeek R1 0508 29.28%
Kimi K2 Thinking 29.23%
GPT-5.1 28%
Gemini 2.5 Flash (Sep) 27%
DeepSeek V3.2 Exp 27%
GLM-4.6 25%
Kimi K2 0905 24%
Llama 4 Maverick 24%
Grok 4.1 Fast 23.50%
Qwen3 235B A22B 2507 22%
MiniMax-M2 21%
Magistral Medium 1.2 20%
gpt-oss-120B (High) 20%
Claude 4.5 Haiku 16%
Llama Nemotron Super 49B v1.5 16%
gpt-oss-20B (High) 15%

Accuracy presents a different picture. Gemini 3 Preview (High) leads the pack at 54 percent, meaning it correctly answers just over half of all questions, followed by Claude Opus 4.5 at 43 percent and Grok 4 at 40 percent. Gemini 2.5 Pro comes next with 37 percent, while GPT-5.1 (High) reaches 35 percent and Claude 4.5 Sonnet 31 percent. A cluster of models then falls into the upper to mid-twenties: DeepSeek R1 0508 at 29.28 percent, Kimi K2 Thinking at 29.23 percent, GPT-5.1 at 28 percent, and both Gemini 2.5 Flash (Sep) and DeepSeek V3.2 Exp at 27 percent. The remaining models descend to GLM-4.6 at 25 percent, Kimi K2 0905 and Llama 4 Maverick at 24 percent, and EXAONE 4.032B at 13 percent. The spread highlights that even the top-performing models answer fewer than six out of ten questions correctly, showing the inherent difficulty AI faces in delivering consistently reliable responses across a broad set of prompts.

Clear Trade-offs

The contrast between hallucination and accuracy charts shows that strong accuracy does not guarantee low hallucination. Some high-ranking models in accuracy still produce incorrect answers at significant rates. Others deliver lower accuracy yet avoid the highest hallucination levels. These gaps illustrate how unpredictable model behavior remains, even as systems improve.

Read next: ChatGPT Doubles Usage as Google Gemini Reaches 40 Percent


by Irfan Ahmad via Digital Information World