New Study Reveals Gaps in Smartwatch's Ability to Detect Undiagnosed High Blood Pressure

In September 2025, the U.S. Food and Drug Administration cleared the Apple Watch Hypertension Notifications Feature, a cuffless tool that uses the watch’s optical sensors to detect blood flow patterns and alert users when their data suggest possible hypertension. While the feature is not intended to diagnose high blood pressure, it represents a step toward wearable-based population screening.

In a new analysis led by investigators from the University of Utah and the University of Pennsylvania and published in the Journal of the American Medical Association, researchers examined what the real-world impact of this technology might look like if deployed broadly across the U.S. adult population.

“High blood pressure is what we call a silent killer,” said Adam Bress, Pharm.D., M.S., senior author and researcher at the Spencer Fox Eccles School of Medicine at the University of Utah. “You can’t feel it for the most part. You don’t know you have it. It’s asymptomatic, and it’s the leading modifiable cause of heart disease.”

How Smartwatches Detect—Or Miss—High Blood Pressure

Apple’s previous validation study found that approximately 59% of individuals with undiagnosed hypertension would not receive an alert, while about 8% of those without hypertension would receive a false alert. Current guidelines recommend using both an office-based blood pressure measurement and an out-of-office blood pressure measurement using a cuffed device to confirm the diagnosis of hypertension. For many people, blood pressure can be different in a doctor’s office compared to their home.

Using data from a nationally representative survey of U.S. adults, Bress and his colleagues estimated how Apple Watch hypertension alerts would change the probability that different populations of adults without a known diagnosis actually have hypertension. The analysis focused on adults aged 22 years or older who were not pregnant and were unaware of having high blood pressure—the population eligible to use the feature.

The analysis revealed important variations: among younger adults under 30, receiving an alert increases the probability of having hypertension from 14% (according to NHANES data) to 47%, while not receiving an alert lowers it to 10%. However, for adults 60 and older—a group with higher baseline hypertension rates—an alert increases the probability from 45% to 81%, while the absence of an alert only lowers it to 34%.

The key takeaway from these data is that as the prevalence of undiagnosed hypertension increases, the likelihood that an alert represents true hypertension also increases. In contrast, the absence of an alert becomes less reassuring as prevalence increases. For example, the absence of an alert is more reassuring in younger adults and substantially less reassuring in older adults and other higher-prevalence subgroups.

The study also found differences across racial and ethnic groups: among non-Hispanic Black adults, receiving an alert increases the probability of having hypertension from 36% to 75%, while not receiving an alert lowers it to 26%. However, for Hispanic adults, an alert increases the probability from 24% to 63%, while its absence lowers the probability to 17%. These differences reflect known disparities in cardiovascular health that are largely driven by social determinants of health, Bress said.

Should You Use Your Smartwatch’s Hypertension Alert Feature?

With an estimated 30 million Apple Watch users in the U.S. and 200 million worldwide, the researchers emphasize that while the notification feature represents a promising public health tool, it should supplement—not replace—standard blood pressure screening with validated cuff-based devices.

“If it helps get people engaged with the health care system to diagnose and treat hypertension using cuff-based measurement methods, that's a good thing,” Bress said.

Current guidelines recommend blood pressure screening every three to five years for adults under 40 and no additional risk factors, and annually for those 40 and older. The researchers caution that false reassurance from not receiving an alert could discourage some individuals from obtaining appropriate cuff-based screening, resulting in missed opportunities for early detection and treatment.

When patients present with an Apple Watch hypertension alert, Bress recommends clinicians perform “a high-quality cuff-based office blood pressure measurement and then consider an out-of-office blood pressure measurement, whether it’s home blood pressure monitoring or ambulatory blood pressure monitoring to confirm the diagnosis.”

The research team plans follow-up studies to estimate the actual numbers of U.S. adults who would receive false negatives and false positives, broken down by region, income, education, and other demographic factors.

The results are published in JAMA as “Impact of a Smartwatch Hypertension Notification Feature for Population Screening.”

The study was supported by the National Heart, Lung, and Blood Institute (R01HL153646) and involved researchers from the University of Utah, the University of Pennsylvania, the University of Sydney, the University of Tasmania, and Columbia University. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Note: This article was originally published by the University of Utah Health Newsroom and is republished here with permission; the Research Communication team confirmed to the DIW team that no AI tools were used in creating the content.

Image: Pexels / Torsten Dettlaff

Read next: YouTubers love wildlife, but commenters aren’t calling for conservation action
by External Contributor via Digital Information World

YouTubers love wildlife, but commenters aren’t calling for conservation action

Edited by Asim BN.

A careful analysis, powered in part by machine learning, highlights an opportunity for conservation messaging on social media

YouTube is a great place to find all sorts of wildlife content. It is not, however, a good place to find viewers encouraging each other to preserve that wildlife, according to new research led by the University of Michigan.

Screenshot: YouTube. Credited DIW.

Out of nearly 25,000 comments posted to more than 1,750 wildlife YouTube videos, just 2% featured a call to action that would help conservation efforts, according to a new study published in the journal Communications Sustainability.

“Our results basically show that people like to watch videos of zoos and safaris and that they appreciate the aesthetics and majesty of certain animals,” said author Derek Van Berkel, associate professor at the U-M School for Environment and Sustainability, or SEAS. “But there really wasn’t much of a nuanced conversation about conservation.”

Although he didn’t expect to see most commenters urging other YouTube users to call their elected officials or to support conservation groups, “I was hoping there might be more,” Van Berkel said. “I thought it might be bigger than 2%.”

Despite the low number, however, the team believes the report still has an optimistic take-home message.

“The flip side of this is we can and should do better at messaging, and there’s a huge potential to do so,” said study co-author Neil Carter, associate professor at SEAS.

While individual YouTube viewers weren’t organically calling for conservation action, there was also a notable absence of conservation groups and influencers working to start conversations and sharing actionable information in the comments.

“There’s tremendous untapped potential for conservation messaging to be improved,” Carter said.

Unlike many other social media platforms, YouTube provided sufficiently accessible, detailed and structured data to provide insights into the digital culture around wildlife conservation, Van Berkel said. And the data was just the starting point.

YouTube’s 8M dataset contained information for nearly 4,000 videos that had been classified as wildlife. The researchers trimmed the list by more than half by selecting videos that featured at least one English language comment and that they could categorize into one of seven topic areas. Those included footage from zoos, safaris and hunting.

The next step was characterizing the comments by the attitudes they expressed. The team arrived at five different categories for these. Expressions of appreciation and concern, both for wildlife and humans, made up four of the categories. The fifth was calls to action.

With the categories and the criteria for each defined, the team created a “gold set” of comment attitudes from 2,778 comments assigned by hand. The researchers then used this data to train a machine learning model to assess more than 20,000 additional comments.

Those steps were painstaking and labor intensive—the team hired additional participants to crowdsource the construction of the comment attitude gold set. But one of the biggest challenges was training the machine learning algorithm on what calls to action looked like when there were so few to begin with, said co-author Sabina Tomkins, assistant professor at the U-M School of Information.

“If the label you’re looking for happens far less often than the others, that problem is really hard. You’re looking for a needle in a haystack,” she said. “The way we solved that challenge was by looking at the models very carefully, figuring out what they were doing.”

Tomkins said the effort from the School of Information graduate students who were part of the research team—Sally Yin, Hongfei Mei, Yifei Zhang and Nilay Gautam—was a driving force behind the project. Enrico Di Minin, a professor at the University of Helsinki, also contributed to the work, which was funded in part by the European Union.

Study: YouTube content on wildlife engages audiences but rarely drives meaningful conservation action (DOI: 10.1038/s44458-025-00018-2).

Contact: Matt Davenport.

Editor’s Notes: This article was originally published on Michigan News, and republished on DIW with permission.

Read next: AI could mark the end of young people learning on the job – with terrible results

by External Contributor via Digital Information World

AI could mark the end of young people learning on the job – with terrible results

Vivek Soundararajan, University of Bath

Image: Tara Winstead / Pexels

For a long time, the deal for a wide range of careers has been simple enough. Entry-level workers carried out routine tasks in return for mentorship, skill development and a clear path towards expertise.

The arrangement meant that employers had affordable labour, while employees received training and a clear career path. Both sides benefited.

But now that bargain is breaking down. AI is automating the grunt work – the repetitive, boring but essential tasks that juniors used to do and learn from.

And the consequences are hitting both ends of the workforce. Young workers cannot get a foothold. Older workers are watching the talent pipeline run dry.

For example, one study suggests that between late 2022 and July 2025, entry-level employment in the US in AI-exposed fields like software development and customer service declined by roughly 20%. Employment for older workers in the same sectors grew.

And that pattern makes sense. AI currently excels at administrative tasks – things like data entry or filing. But it struggles with nuance, judgment and plenty of other skills which are hard to codify.

So experience and the accumulation of those skills become a buffer against AI displacement. Yet if entry-level workers never get the chance to build that experience, the buffer never forms.

This matters for organisations too. Researchers using a huge amount of data about work in the US described the way that professional skills develop over time, by likening career paths to the structure of a tree.

General skills (communication, critical thinking, problem solving) form the trunk, and then specialised skills branch out from there.

Their key finding was that wage premiums for specialised skills depend almost entirely on having those strong general foundational skills underneath. Communication and critical thinking capabilities are not optional extras – they are what make advanced skills valuable.

The researchers also found that workers who lack access to foundational skills can become trapped in career paths with limited upward mobility: what they call “skill entrapment”. This structure has become more pronounced over the past two decades, creating what the researchers described as “barriers to upward job mobility”.

But if AI is eliminating the entry-level positions where those foundations were built, who develops the next generation of experts? If AI can do the junior work better than the actual juniors, senior workers may stop delegating altogether.

Researchers call this a “training deficit”. The junior never learns, and the pipeline breaks down.

Uneven disruption

But the disruption will not hit everyone equally. It has been claimed, for example, that women face nearly three times the risk of their jobs being replaced with AI compared to men.

This is because women are generally more likely to be in clerical and administrative roles, which are among the most exposed to AI-driven transformation. And if AI closes off traditional routes into skilled work, the effects are unlikely to be evenly distributed.

So what can be done? Well, just because the old pathway deal between junior and senior human workers is broken, does not mean that a new one cannot be built.

Young workers now need to learn what AI cannot replace in terms of knowledge, judgment and relationships. They need to seek (and be provided with) roles which involve human interaction, rather than just screen-based tasks. And if traditional entry-level jobs are disappearing, they need to look for structured programmes that still offer genuine skill development.

Older workers meanwhile, can learn a lot from younger workers about AI and technology. The idea of mentorship can be flipped, with juniors teaching about new tools, while seniors provide guidance and teaching on nuance and judgment.

And employers need to resist the urge to cut out junior staff. They should keep delegating to those staff – even when AI can do the job more quickly. Entry level roles can be redesigned rather than eliminated. For ultimately, if juniors are not getting trained, there will be no one to hand over to.

Protecting the pipeline of skilled and valuable employees is in everyone’s interest. Yes, some forms of expertise will matter less in the age of AI, which is disorienting for people who may have invested years in developing them.

But expertise is not necessarily about storing information. It is also about refined judgment being applied to complex situations. And that remains valuable.

Vivek Soundararajan, Professor of Work and Equality, University of Bath

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

Read next: Could LLMs Repeat False Medical Claims When They Are Confidently Worded? Study Reports They Can

by External Contributor via Digital Information World

Could LLMs Repeat False Medical Claims When They Are Confidently Worded? Study Reports They Can

Edited by Asim BN.

Medical artificial intelligence (AI) is often described as a way to make patient care safer by helping clinicians manage information. A new study by the Icahn School of Medicine at Mount Sinai and collaborators confronts a critical vulnerability: when a medical lie enters the system, can AI pass it on as if it were true?

Image: Enchanted Tools / Unsplash

Analyzing more than a million prompts across nine leading language models, the researchers found that these systems can repeat false medical claims when they appear in realistic hospital notes or social-media health discussions.

The findings, published in the February 9 online issue of The Lancet Digital Health [10.1016/j.landig.2025.100949], suggest that current safeguards do not reliably distinguish fact from fabrication once a claim is wrapped in familiar clinical or social-media language.

To test this systematically, the team exposed the models to three types of content: real hospital discharge summaries from the Medical Information Mart for Intensive Care (MIMIC) database with a single fabricated recommendation added; common health myths collected from Reddit; and 300 short clinical scenarios written and validated by physicians. Each case was presented in multiple versions, from neutral wording to emotionally charged or leading phrasing similar to what circulates on social platforms.

In one example, a discharge note falsely advised patients with esophagitis-related bleeding to “drink cold milk to soothe the symptoms.” Several models accepted the statement rather than flagging it as unsafe. They treated it like ordinary medical guidance.

“Our findings show that current AI systems can treat confident medical language as true by default, even when it’s clearly wrong,” says co-senior and co-corresponding author Eyal Klang, MD, Chief of Generative AI in the Windreich Department of Artificial Intelligence and Human Health at the Icahn School of Medicine at Mount Sinai. “A fabricated recommendation in a discharge note can slip through. It can be repeated as if it were standard care. For these models, what matters is less whether a claim is correct than how it is written.”

The authors say the next step is to treat “can this system pass on a lie?” as a measurable property, using large-scale stress tests and external evidence checks before AI is built into clinical tools.

“Hospitals and developers can use our dataset as a stress test for medical AI,” says physician-scientist and first author Mahmud Omar, MD, who consults with the research team. “Instead of assuming a model is safe, you can measure how often it passes on a lie, and whether that number falls in the next generation.”

“AI has the potential to be a real help for clinicians and patients, offering faster insights and support,” says co-senior and co-corresponding author Girish N. Nadkarni, MD, MPH, Chair of the Windreich Department of Artificial Intelligence and Human Health, Director of the Hasso Plattner Institute for Digital Health, Irene and Dr. Arthur M. Fishberg Professor of Medicine at the Icahn School of Medicine at Mount Sinai, and Chief AI Officer of the Mount Sinai Health System. “But it needs built-in safeguards that check medical claims before they are presented as fact. Our study shows where these systems can still pass on false information, and points to ways we can strengthen them before they are embedded in care.”

The paper is titled “Mapping LLM Susceptibility to Medical Misinformation Across Clinical Notes and Social Media.” 

The study’s authors, as listed in the journal, are Mahmud Omar, Vera Sorin, Lothar H Wieler, Alexander W Charney, Patricia Kovatch, Carol R Horowitz, Panagiotis Korfiatis, Benjamin S Glicksberg, Robert Freeman, Girish N Nadkarni, and Eyal Klang.

This work was supported by the Clinical and Translational Science Awards (CTSA) grant UL1TR004419 from the National Center for Advancing Translational Sciences. Research reported in this publication was also supported by the Office of Research Infrastructure of the National Institutes of Health under award number S10OD026880 and S10OD030463.

For more Mount Sinai artificial intelligence news, visit: https://icahn.mssm.edu/about/artificial-intelligence.

Note: This post was originally published on Mount Sinai and republished on DIW with permission.

Read next:

• Study: Platforms that rank the latest LLMs can be unreliable

• Google Expands ‘Results About You’ Tool to Include Government ID Monitoring

by Press Releases via Digital Information World

Study: Platforms that rank the latest LLMs can be unreliable

Removing just a tiny fraction of the crowdsourced data that informs online ranking platforms can significantly change the results.

By Adam Zewe | MIT News.

Image: Markus Winkler / Pexels

A firm that wants to use a large language model (LLM) to summarize sales reports or triage customer inquiries can choose between hundreds of unique LLMs with dozens of model variations, each with slightly different performance.

To narrow down the choice, companies often rely on LLM ranking platforms, which gather user feedback on model interactions to rank the latest LLMs based on how they perform on certain tasks.

But MIT researchers found that a handful of user interactions can skew the results, leading someone to mistakenly believe one LLM is the ideal choice for a particular use case. Their study reveals that removing a tiny fraction of crowdsourced data can change which models are top-ranked.

They developed a fast method to test ranking platforms and determine whether they are susceptible to this problem. The evaluation technique identifies the individual votes most responsible for skewing the results so users can inspect these influential votes.

The researchers say this work underscores the need for more rigorous strategies to evaluate model rankings. While they didn’t focus on mitigation in this study, they provide suggestions that may improve the robustness of these platforms, such as gathering more detailed feedback to create the rankings.

The study also offers a word of warning to users who may rely on rankings when making decisions about LLMs that could have far-reaching and costly impacts on a business or organization.

“We were surprised that these ranking platforms were so sensitive to this problem. If it turns out the top-ranked LLM depends on only two or three pieces of user feedback out of tens of thousands, then one can’t assume the top-ranked LLM is going to be consistently outperforming all the other LLMs when it is deployed,” says Tamara Broderick, an associate professor in MIT’s Department of Electrical Engineering and Computer Science (EECS); a member of the Laboratory for Information and Decision Systems (LIDS) and the Institute for Data, Systems, and Society; an affiliate of the Computer Science and Artificial Intelligence Laboratory (CSAIL); and senior author of this study.

She is joined on the paper by lead authors and EECS graduate students Jenny Huang and Yunyi Shen as well as Dennis Wei, a senior research scientist at IBM Research. The study will be presented at the International Conference on Learning Representations.

Dropping data

While there are many types of LLM ranking platforms, the most popular variations ask users to submit a query to two models and pick which LLM provides the better response.

The platforms aggregate the results of these matchups to produce rankings that show which LLM performed best on certain tasks, such as coding or visual understanding.

By choosing a top-performing LLM, a user likely expects that model’s top ranking to generalize, meaning it should outperform other models on their similar, but not identical, application with a set of new data.

The MIT researchers previously studied generalization in areas like statistics and economics. That work revealed certain cases where dropping a small percentage of data can change a model’s results, indicating that those studies’ conclusions might not hold beyond their narrow setting.

The researchers wanted to see if the same analysis could be applied to LLM ranking platforms.

“At the end of the day, a user wants to know whether they are choosing the best LLM. If only a few prompts are driving this ranking, that suggests the ranking might not be the end-all-be-all,” Broderick says.

But it would be impossible to test the data-dropping phenomenon manually. For instance, one ranking they evaluated had more than 57,000 votes. Testing a data drop of 0.1 percent means removing each subset of 57 votes out of the 57,000, (there are more than 10194subsets), and then recalculating the ranking.

Instead, the researchers developed an efficient approximation method, based on their prior work, and adapted it to fit LLM ranking systems.

“While we have theory to prove the approximation works under certain assumptions, the user doesn’t need to trust that. Our method tells the user the problematic data points at the end, so they can just drop those data points, re-run the analysis, and check to see if they get a change in the rankings,” she says.

Surprisingly sensitive

When the researchers applied their technique to popular ranking platforms, they were surprised to see how few data points they needed to drop to cause significant changes in the top LLMs. In one instance, removing just two votes out of more than 57,000, which is 0.0035 percent, changed which model is top-ranked.

A different ranking platform, which uses expert annotators and higher quality prompts, was more robust. Here, removing 83 out of 2,575 evaluations (about 3 percent) flipped the top models.

Their examination revealed that many influential votes may have been a result of user error. In some cases, it appeared there was a clear answer as to which LLM performed better, but the user chose the other model instead, Broderick says.

“We can never know what was in the user’s mind at that time, but maybe they mis-clicked or weren’t paying attention, or they honestly didn’t know which one was better. The big takeaway here is that you don’t want noise, user error, or some outlier determining which is the top-ranked LLM,” she adds.

The researchers suggest that gathering additional feedback from users, such as confidence levels in each vote, would provide richer information that could help mitigate this problem. Ranking platforms could also use human mediators to assess crowdsourced responses.

For the researchers’ part, they want to continue exploring generalization in other contexts while also developing better approximation methods that can capture more examples of non-robustness.

“Broderick and her students’ work shows how you can get valid estimates of the influence of specific data on downstream processes, despite the intractability of exhaustive calculations given the size of modern machine-learning models and datasets,” says Jessica Hullman, the Ginni Rometty Professor of Computer Science at Northwestern University, who was not involved with this work. “The recent work provides a glimpse into the strong data dependencies in routinely applied — but also very fragile — methods for aggregating human preferences and using them to update a model. Seeing how few preferences could really change the behavior of a fine-tuned model could inspire more thoughtful methods for collecting these data.”

This research is funded, in part, by the Office of Naval Research, the MIT-IBM Watson AI Lab, the National Science Foundation, Amazon, and a CSAIL seed award.

Paper: "Dropping Just a Handful of Preferences Can Change Top Large Language Model Rankings".

Reprinted with permission of MIT News.

Read next:

• Why comparisons between AI and human intelligence miss the point

• Google Expands ‘Results About You’ Tool to Include Government ID Monitoring

by External Contributor via Digital Information World

Google Expands ‘Results About You’ Tool to Include Government ID Monitoring

Reviewed By Asim BN.

Google announced on Feb. 10, 2026 that it is expanding its “Results about you” tool to help users find and request the removal of Search results containing government-issued identification numbers.

In a blog post, Product Manager Phoebe Wong said, “Over 10 million people have used the ‘Results about you’ tool to control how their sensitive personal information appears online, like a phone number or home address.” The company added that users can now find and request the removal of information such as “your driver’s license, passport, or Social Security number.”

Users can access the tool in the Google app by clicking their account photo and selecting “Results about you,” or by visiting goo.gle/resultsaboutyou. First-time users are prompted to “add the personal contact information you want to monitor,” including government ID numbers, while existing users can add ID numbers directly.

Image: Google / The Keyword Blog

Google said the tool “employs Google’s rigorous security protocols and advanced encryption to prevent misuse and ensure your privacy.” Once confirmed, the system automatically monitors Search results and notifies users if matches are found.

The company noted, “Removing this information from Google Search doesn’t remove it from the web entirely,” and said the update will roll out in the United States first, with plans for additional regions.

Google’s FAQ on the “Results about you” tool emphasizes that users’ personal information is handled with strict security standards. The company states, “We take that responsibility very seriously. To prevent misuse, we store your personal info in accordance with Google's high standards for sensitive personally identifiable information, which includes advanced encryption and access controls.”

The information users provide, such as phone numbers, home addresses, email addresses, and government-issued IDs, is used only for monitoring, to process removal requests, and to improve the monitoring and removal process, and is not shared across other Google products or used for advertising.

Although Google’s FAQ does not specifically address legal disclosures, information submitted through the tool could be provided to authorities under judicial oversight or other legally binding government requests, as confirmed by the DIW team from Google’s published privacy policy.

The company also warns that misuse of the tool may result in losing access or other consequences under its Terms of Service.

Notes: This post was improved with the assistance of AI tools.

Read next: When both partners work from home: the hidden cost of always-on technology

by Ayaz Khan via Digital Information World

When both partners work from home: the hidden cost of always-on technology

By Craig Donaldson. Edited by Asim BN.

Image: Jack Sparrow / Pexels

When partners work from home, constant digital interruptions increase after-work frustration, strain couples’ relationships and place a heavier psychological burden on women, UNSW research has found.

Work-from-home couples experience heightened frustration and relationship conflict when technology allows the intrusion of work into family time, according to new research from UNSW Business School.

Furthermore, women bear a disproportionate psychological burden from these digital interruptions, though simple planning strategies can help reduce the negative impacts of work technology on home life for couples.

The research examined what academics call ICT permeability, which describes how information and communication technologies such as email, text messaging, mobile phones and remote meeting applications pierce the once-solid barrier between work and home life.

When both partners work from home, this blurring of boundaries was associated with distinct challenges that differed markedly from households where only one partner worked remotely.

Understanding the research behind remote work challenges

Manju Ahuja, Scientia Professor in the School of Information Systems and Technology Management at UNSW Business School, investigated these challenges together with her co-authors in a comprehensive study, Work-Family Frustration When You and Your Partner Both Work from Home: The Role of ICT Permeability, Planning, and Gender.

Published in the Journal of the Association for Information Systems, the study is based on a 10-day diary study with 117 participants who lived with their partners while both worked from home full-time through the COVID-19 pandemic.

Participants responded to three daily online surveys over consecutive workdays, providing real-time insights into their experiences. This methodology enabled researchers to capture daily, real-time fluctuations in work-family dynamics and to reduce the retrospective bias common in traditional surveys.

This new study follows up on Prof. Ahuja’s previous research on the psychological and relational costs of working from home, despite the benefits of flexibility and avoiding the daily commute. “This overall stream of research explores the double-edged sword of technology (such as Zoom and Teams) and anytime-anywhere connectivity,” Prof. Ahuja explained.

“The previous research found that, while employees reported significantly improved productivity, they also tended to suffer from stress-related physiological symptoms (like headaches), and their relationships were adversely affected.

“With this new study, we wanted to examine whether these effects are exacerbated when both partners work from home. We were trying to understand what employees can do if they wish to maintain some form of work-life balance in the face of relentless connectivity and constant negotiations of home and work tasks with their partners.”

When work never stops

The study found that work-related technology use during personal time was associated with depletion of individuals’ limited cognitive and emotional resources, leading to what researchers termed “after-work frustration”. This frustration reflected the negative emotions people experienced when unable to fulfil family activities or personal responsibilities due to work-related interruptions.

The research also revealed a counterintuitive finding about productivity. After-work frustration was positively associated with increased job productivity in the short term, as individuals redirected their limited resources towards work tasks where they perceived higher likelihood of success and recognition.

However, this productivity boost came at a cost to family relationships. As the study noted, “attempting to complete family tasks while facing work-related activities is likely to induce frustration.”

“It is important to discuss the ways in which remote work affects work productivity in a meaningful and nuanced manner,” said Prof. Ahuja. “While it is important to look at the short-term productivity gains, it is equally important to look at the effects on the overall lives of employees because personal and professional get quite entangled in remote work settings, and eventually affect work outcomes.”

A gender divide emerges

Women experienced significantly stronger negative effects from technology-enabled work intrusions than men. The research demonstrated that despite increased workforce participation and evolving gender roles, women continued to shoulder greater responsibility for domestic chores, childcare and relationship maintenance.

The study found women are more adversely impacted for not fulfilling family demands, while men traditionally faced greater pressure to meet work obligations. This dynamic meant technology intrusions during family time were associated with more psychological distress for women.

The researchers observed that “women are often responsible for invisible labour – unnoticed and undervalued work at home that includes household chores, childcare, and emotional support for family members.”

When both partners worked from home simultaneously, neither could provide the buffering support typically available when only one partner worked remotely. Each partner faced their own technology intrusions while also managing their partner’s work demands, creating compound stress that intensified frustration levels.

Planning associated with reduced frustration

The research identified daily planning as a strategy associated with reduced frustration from technology-enabled work intrusions. Planning behaviour, which involved setting goals and prioritising tasks for both work and family activities, was linked to helping individuals maintain clearer boundaries between professional and personal domains.

The study also found that individuals who engaged in structured planning experienced weaker relationships between technology permeability and after-work frustration. Planning was associated with more effective resource allocation, reducing the sense of being overwhelmed by competing demands. The research showed that “planning behaviour is particularly salient in the context of working from home.”

“This suggests that when the remote-working partners engage in joint daily planning to account for meetings and video calls each has scheduled (which can be problematic in certain home office setups) and the domestic and childcare tasks that need to be accomplished at certain times, they face lower levels of frustration with each other and internally,” said Prof. Ahuja.

Practical implications for workers and organisations

For individuals, the research suggested actionable strategies. Couples can establish technology-free times and zones, such as a daily hour during dinner and family activities. Sharing schedules between partners allow for better coordination and reduced miscommunication about availability. Digital tools like shared calendars can also help manage the boundary between work and family time.

Organisations also have a role to play in supporting work-from-home arrangements. Employees can be trained in planning techniques to enhance time management skills and reduce family conflict.

Similarly, Prof. Ahuja said managers need to be trained in supervising remote workers and workplace cultures that accommodate home and relational life should be cultivated. “For example, employers can allow employees to block out personal time on shared calendars can also help reduce interruptions during family activities,” she said.

“The research recommended developing policies that granted employees control over flexible work while establishing clear boundaries. These policies could include restricting non-urgent work meetings after 6pm or during weekends and encouraging employees to communicate their availability within teams.”

“These proactive remote work management strategies are likely to lead to higher satisfaction of remote workers, leading to higher employee retention,” Prof. Ahuja concluded.

Note: This post was originally published on UNSW (The University of New South Wales) Sydney Newsroom and republished on DIW with permission.

Read next:

• Why comparisons between AI and human intelligence miss the point

• Why The Real Cost of Working From Home Varies Wildly in US Cities

by External Contributor via Digital Information World