Health
Researchers Develop AI Method That Makes Computer Vision Models More Explainable
A new technique developed by MIT researchers could help make artificial intelligence systems more accurate and transparent in high-stakes fields such as health care and autonomous driving by improving how computer vision models explain their decisions.
MIT researchers have developed a new explainable AI method that improves the accuracy and transparency of computer vision models, helping users trust AI predictions in healthcare and autonomous driving.
Researchers at MIT have developed a new approach to make computer vision models more transparent, offering a potential boost to trust and accountability in safety-critical applications such as medical diagnosis and autonomous driving.
In a media statement, the researchers said the method improves on a widely used explainability technique known as concept bottleneck modeling, which enables AI systems to show the human-understandable concepts behind a prediction. The new approach is designed to produce clearer explanations while also improving prediction accuracy.
Why explainable AI matters
In areas such as health care, users often need more than just a model’s output. They want to understand why a system arrived at a particular conclusion before deciding whether to rely on it. Concept bottleneck models attempt to address that need by forcing an AI system to make predictions through a set of intermediate concepts that humans can interpret.
For example, when analysing a medical image for melanoma, a clinician might define concepts such as “clustered brown dots” or “variegated pigmentation.” The model would first identify those concepts and then use them to arrive at its final prediction.
But the researchers said pre-defined concepts can sometimes be too broad, irrelevant or incomplete for a specific task, limiting both the quality of explanations and the model’s performance. To overcome that, the MIT team developed a method that extracts concepts the model has already learned during training and then compels it to use those concepts when making decisions.
The approach relies on two specialised machine-learning models. One extracts the most relevant internal features learned by the target model, while the other translates them into plain-language concepts that humans can understand. This makes it possible to convert a pretrained computer vision model into one capable of explaining its reasoning through interpretable concepts.
“In a sense, we want to be able to read the minds of these computer vision models. A concept bottleneck model is one way for users to tell what the model is thinking and why it made a certain prediction. Because our method uses better concepts, it can lead to higher accuracy and ultimately improve the accountability of black-box AI models,” Antonio De Santis, lead author of the study, said in a media statement.
De Santis is a graduate student at Polytechnic University of Milan and carried out the research while serving as a visiting graduate student at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL). The paper was co-authored by Schrasing Tong, Marco Brambilla of Polytechnic University of Milan, and Lalana Kagal of CSAIL. The research will be presented at the International Conference on Learning Representations.
Concept bottleneck models have gained attention as a way to improve AI explainability by introducing an intermediate reasoning step between an input image and the final output. In one example, a bird-classification model might identify concepts such as “yellow legs” and “blue wings” before predicting a barn swallow.
However, the researchers noted that these concepts are often generated in advance by humans or large language models, which may not always match the needs of the task. Even when a model is given a fixed concept set, it can still rely on hidden information not visible to users, a challenge known as information leakage.
“These models are trained to maximize performance, so the model might secretly use concepts we are unaware of,” De Santis said in a media statement.
The team’s solution was to tap into the knowledge the model had already acquired from large volumes of training data. Using a sparse autoencoder, the method isolates the most relevant learned features and reconstructs them into a small number of concepts. A multimodal large language model then describes each concept in simple language and labels the training images by marking which concepts are present or absent.
The annotated dataset is then used to train a concept bottleneck module, which is inserted into the target model. This forces the model to make predictions using only the extracted concepts.
The researchers said one of the biggest challenges was ensuring that the automatically identified concepts were both accurate and understandable to humans. To reduce the risk of hidden reasoning, the model is limited to just five concepts for each prediction, encouraging it to focus only on the most relevant information and making the explanation easier to follow.
When tested against state-of-the-art concept bottleneck models on tasks including bird species classification and skin lesion identification, the new method delivered the highest accuracy while also producing more precise explanations, according to the researchers. It also generated concepts that were more relevant to the images in the dataset.
Still, the team acknowledged that the broader challenge of balancing accuracy and interpretability remains unresolved.
“We’ve shown that extracting concepts from the original model can outperform other CBMs, but there is still a tradeoff between interpretability and accuracy that needs to be addressed. Black-box models that are not interpretable still outperform ours,” De Santis said in a media statement.
Looking ahead, the researchers plan to explore ways to further reduce information leakage, possibly by adding additional concept bottleneck modules. They also aim to scale up the method by using a larger multimodal language model to annotate a larger training dataset, which could improve performance further.
This latest work adds to growing efforts to make AI systems not only more powerful, but also more understandable in domains where trust can be as important as accuracy.
Climate
When the Monsoon Returns, Fear Returns: How Climate Disasters Are Reshaping Mental Health in Himachal Pradesh
In Himachal Pradesh, repeated floods and landslides are reshaping mental health in the mountains — but disaster policy still doesn’t see it.
“As soon as the monsoon arrives, fear spreads across the village. People panic because they have already seen what these rains can do. In 2023 and again in 2025, many families here lost their homes, farmland and orchards to monsoon disasters. The moment it starts raining heavily, everyone fears history will repeat itself.”
Het Ram, a resident of Siraj Valley in Himachal Pradesh’s Mandi district, pauses before adding a detail that has quietly become part of everyday life. Most households in his village now keep their important documents, first-aid kits, clothes, dry food, drinking water, ropes and tarpaulin packed together in one corner of the house, ready to be carried at a moment’s notice if disaster strikes.
“Without our documents, we cannot even claim government compensation,” he says.
Why Himachal Pradesh Floods Continue to Haunt Survivors
Het Ram’s experience is shared by families across many villages in Himachal Pradesh, where the monsoon — once welcomed as a season of abundance — has increasingly become synonymous with fear. With the first spell of July rain comes an unspoken anxiety that settles over entire communities. People’s eyes remain fixed not on the clouds but on the fragile mountain slopes towering above their homes. A spell of intense rainfall is enough to keep families awake through the night. Fluctuating blood pressure, panic attacks, insomnia and chronic stress have gradually become a routine part of life in the state’s vulnerable mountain settlements.

Psychologists describe this condition as anticipatory anxiety — the distress that emerges not after a traumatic event but from the constant fear that it may happen again. A study of disaster survivors in Uttarakhand found that repeated exposure to natural disasters leaves people in a permanent state of hypervigilance: interrupted sleep, startling at the slightest sound, and feeling anxious the moment it begins to rain had become routine for many of those affected.
‘I fear the mountain above my home may collapse’

For 56-year-old Chudamani of Sarpada village in Rampur subdivision, July is no longer just another month — it is a season of dread.
Whenever it rains, sleep deserts her. She spends hours sitting outside her house, her eyes fixed on the hillside looming overhead. A single fear keeps returning: What if the mountain gives way? What if everything is gone within seconds?
The fear has haunted her ever since the devastating flash floods of July 2024, when a swollen Samej Khad swept away both her two-storey house and her vehicle.
“Nowhere feels safe anymore,” says her husband, Chatar Singh. “The moment the rains begin, fear settles into our minds. Every spell of heavy rain makes us wonder which family’s home will be destroyed next.”
The scale of destruction behind the anxiety
According to the Himachal Pradesh State Disaster Management Authority, 2,108 people lost their lives during the monsoon seasons between 2020 and 2025, while property worth thousands of crores of rupees was destroyed.
The 2023 monsoon alone claimed 404 lives, left 38 people missing and forced more than 50,000 people to abandon their homes. A total of 2,546 houses were completely destroyed and another 10,853 suffered partial damage. Overall losses crossed INR 12,000 crore.

On August 14, 2023, a landslide in Shimla city killed 20 people. The following year, a cloudburst in the Samej area of Shimla district triggered flash floods that washed away 32 houses and claimed 36 lives. These recurring disasters across Shimla, Mandi and Kullu districts have left behind more than physical and economic devastation — every monsoon now revives memories of loss, uncertainty and the fear of losing loved ones all over again.
‘Not even a needle found after floods’
Alok Kumar of Sarpada village in Rampur embodies the depth of this trauma.
His ancestral home in Samej village was completely washed away in the 2024 floods. His mother and eight tenants lost their lives in the disaster.
“The flood was so devastating that not even a needle remained from our 21-room house,” he recalls. “I still haven’t recovered from that tragedy. Every time I hear about floods or landslides, the entire scene flashes before my eyes. We live in the mountains, so there is always the fear that another disaster could strike at any time.”
He says disasters have become an annual reality.
“For the last three years, something has happened every monsoon. It has increased stress and anxiety among almost everyone in the village. Even the elderly say they have never witnessed anything like this before.”
Despite struggling emotionally, Alok has never sought professional help. Like many others in disaster-prone regions, he considers such distress an unavoidable part of life in the hills.
What doctors are seeing
Mental health professionals in Himachal Pradesh say the psychological toll of climate disasters is becoming more visible, even though systematic evidence from the state is still lacking.
Dr Dinesh Dutt, a psychiatrist at Indira Gandhi Medical College (IGMC), Shimla, says climate change is directly influencing people’s mental well-being.
“Climate has a direct impact on mental health. After monsoon disasters, many survivors continue to relive the traumatic event. Distressing thoughts, memories and images keep recurring in their minds. Clinically, we recognise these symptoms as anxiety and post-traumatic stress disorder (PTSD). We are seeing such cases in hospitals, and our assessments suggest that climate-related events are contributing to these psychological conditions. However, there is still no systematic research that conclusively establishes this link in Himachal Pradesh.”
He adds that the state currently lacks a comprehensive database documenting climate-related mental health cases.
The concerns raised by clinicians are echoed in scientific literature. A review published in Frontiers in Public Health (2025) found that sudden climate-related disasters such as floods, landslides and cyclones frequently lead to PTSD and acute anxiety, while repeated exposure to such events has been associated with depression, hopelessness and, in some cases, suicidal thoughts. Mental health experts believe a similar pattern is gradually emerging in the rural regions of Himachal Pradesh.
Children are carrying the burden too
The psychological impact of recurring disasters is becoming increasingly visible among children.
Mukta, a government schoolteacher in Rohru, says classroom behaviour has changed noticeably since the devastating monsoon of 2023.
“Children have become frightened of the rains. During heavy rainfall, some of them become unusually quiet. Since Rohru is an apple-growing region, many children also worry that their family’s orchards and livelihoods may be destroyed.”
Poonam’s ten-year-old daughter, who lives in Shimla, has also developed an intense fear of the monsoon. Their residential building was declared unsafe after a landslide last year, forcing the family to live with relatives for ten days.
“Since that incident, she has developed a deep sense of insecurity,” Poonam says. “Every time it starts raining, that fear returns.”
For teenagers, the consequences are often less visible but equally profound.
Ashish Dogra, a Class XII student from Ramnagar in Shimla, says the rainy season has transformed childhood routines.
“During the monsoon, no one goes out to play anymore. We stay indoors all day. We talk to friends on the phone, but after a while it becomes boring. Sometimes that isolation makes us irritable and angry.”
Psychologist Neelam Bali says children’s emotional responses are closely linked to the anxiety they observe within their own families.
“When children see their parents constantly worried, they internalise that stress. Most are unable to express their emotions openly, and over time the anxiety begins to show in their behaviour. It eventually affects their studies and overall school performance.”
She emphasises that parents need to encourage open conversations with children.
“Play is one of the most effective ways for children to cope with stress. If they remain cut off from outdoor activities for long periods, they are more likely to experience loneliness, irritability and anger.”
Dr Dutt adds that climate disasters affect children, adults and older people in different ways. “If survivors do not receive timely psychological first aid, the long-term health consequences can be significant.”
When landscapes become a source of grief
Natural disasters alter far more than homes, roads and livelihoods. They also transform people’s emotional relationship with the landscapes they have known all their lives.
When the mountains that once symbolised safety and belonging suddenly become a source of destruction, recovering emotionally is rarely easy.
Researchers describe this experience as solastalgia — the distress people feel when the environment they identify as home is damaged or transformed. Unlike nostalgia, which is a longing for a place left behind, solastalgia emerges while people remain in the very place they call home, watching it deteriorate around them. Over time, this distress can deepen into prolonged grief, anxiety and depression.
Shifting homes in search of safety
Discussions around migration from Himachal Pradesh’s hill districts have traditionally centred on employment, education and infrastructure. But a new driver is quietly emerging: the search for psychological security.
For some families, the fear of living in disaster-prone areas has become so overwhelming that they are abandoning ancestral homes for places they perceive to be safer.
Prabhat Chand, 47, a government school lecturer from Rampur, is one such resident. After the devastating floods in Samej, he decided to leave the area and build a new home nearly 17 kilometres away in Sarpada village.

“Most of the families affected by the disaster have moved elsewhere,” he says. “The flood was so destructive that the place no longer feels fit for habitation.”
Ruchi, who lives in Tutu on the outskirts of Shimla, also shifted to another rented house after repeated monsoon scares. Her previous home stood beside a seasonal stream, and every spell of heavy rainfall made her anxiety worse.
“Whenever I saw news reports of rain-related disasters, I became anxious. Watching the water level in the stream rise only intensified that fear,” she says.
Although she still worries during the monsoon, moving to a different neighbourhood has eased the constant panic she once experienced.
Mental health remains missing from disaster response
Himachal Pradesh has made significant progress in recognising the health risks posed by climate change. The state has adopted the State Action Plan on Climate Change and Human Health (2022–27) and established a Climate Change and Human Health Cell to strengthen public health preparedness.

The plan focuses on physical health, disease surveillance, healthcare infrastructure resilience, early warnings and logistics, but does not address mental health support or psychological care for disaster-affected populations.
While the Climate Change and Human Health Cell provides training to healthcare workers and Accredited Social Health Activists (ASHAs) on mental health and psychological first aid, experts say these efforts have yet to translate into meaningful support on the ground.
Dr Sanjay Pathak, former Director of the Himachal Pradesh Mental Health Authority, believes disaster management continues to prioritise physical rehabilitation over psychological recovery.
“Government responses after disasters understandably focus on relief and rehabilitation,” he says. “But very little attention is paid to what kind of psychological intervention survivors require. Unless mental health becomes an integral part of disaster response, people will continue to live with a persistent sense of insecurity.”
Dr Dutt points to another challenge: the severe shortage of trained mental health professionals.
“The frequency of disasters is increasing much faster than our capacity to provide psychological support. For now, community and family networks are helping people cope. But as those social structures weaken, the need for professional mental health services will become much greater.”
The policy gap extends beyond Himachal Pradesh. An analysis published by the Observer Research Foundation (ORF) notes that disaster management strategies in India remain largely centred on rescue operations, rehabilitation and rebuilding infrastructure, with mental health support yet to be systematically integrated into these frameworks. The analysis also found that psychological disorders following natural disasters are 40 per cent more common than physical injuries — underscoring the need to treat mental health as a core part of disaster preparedness and recovery.
What needs to change?
Both Dr Pathak and Dr Dutt agree that psychosocial care must become a standard part of emergency relief rather than an afterthought. They identify three immediate priorities: training schoolteachers to provide basic psychosocial support to children affected by disasters; strengthening the capacity of ASHA workers and frontline health personnel to identify and respond to psychological distress; and expanding access to counselling and mental health services in disaster-affected communities.
For a Himalayan state facing ever more frequent extreme weather, integrating mental health into disaster management is no longer simply a public health recommendation — it is a necessity.
As climate change continues to reshape the mountains, it is also reshaping the emotional landscape of the people who call them home. The landslides and floods may disappear once the rains recede, but for many survivors, the fear remains — returning every monsoon, long before the first cloudburst.
Raman Kant is a mentee at the Asian College of Journalism’s Climate Change Media Hub, supported by Interlink Academy, Germany.
Health
What Forests Could Hold for the Future of Cancer Treatment
Cancer treatment has benefited from plant-derived compounds, including medicines used against leukaemia, lymphoma, breast and ovarian cancers. But deforestation could destroy plants that may contain compounds with potential for future drug discovery.
Cancer remains one of the biggest challenges in medicine. While treatments have improved significantly, there is no single cure for cancer, and some cancers remain difficult to treat. Researchers continue to look for drugs that can target cancer cells more effectively and overcome treatment resistance.
cancer treatment medicines used today were developed from compounds found in plants. Researchers are still studying plants for molecules that could be developed into new drugs. But deforestation could reduce the number of species available for this research, a new analysis suggests.
An analysis by Zero Carbon Analytics estimates that tropical forests contain about USD 714 billion in potential commercial pharmaceutical discovery value from flowering plants. It estimates that deforestation between 2001 and 2024 has already put about USD 86 billion of this potential value at risk. The finding is relevant to cancer research because natural products have contributed to several cancer medicines.
Cancer Medicines From Plants
The link between forests and cancer treatment is already well established. Vincristine and vinblastine, derived from the Madagascar periwinkle, are used against leukaemia, lymphoma and other cancers. Paclitaxel, first isolated from the Pacific yew, is used to treat cancers including breast and ovarian cancer. Camptothecin, from Camptotheca acuminata, led to the development of the cancer drugs irinotecan and topotecan.
Their contribution is part of a wider pattern. A review found that about 25% of anticancer drugs approved between 1981 and 2019 were related to natural products. Zero Carbon Analytics reports that three in four cancer drugs discovered between 1940 and 2014 were natural products, derived from them or inspired by them.
The Research Gap
The bigger concern is what forests could still offer cancer treatment research in the future. The analysis calls this “option value” — the value of keeping open the possibility of discovering new medicines from plants that have not yet been studied.
That possibility is large. The analysis estimates that 99% of plant compounds remain unexplored and that tropical forests could contain around 328 plant-derived drugs with potential commercial value, although the estimate ranges from 200 to 460.
These are not 328 ready-made cancer drugs. Most compounds found in plants will not become medicines and must first go through laboratory tests, safety checks and clinical trials.

But once a species is lost before scientists study it, that opportunity may be lost too. For cancer research, deforestation could therefore mean losing potential sources of new cancer treatments before researchers know what they contain.
What This Means For The Western Ghats?
India’s Western Ghats shows how much remains to be studied. A 2025 review recorded 1,628 medicinal plant species used traditionally in the region, but only 115 had been validated through scientific studies or clinical use.
Some are already being tested for possible cancer-related properties. A 2025 study of Tarenna alpestris, a plant from the Megamalai Hills, found anticancer activity against colon cancer cells in the laboratory.
This does not mean the plant can treat cancer. More research is needed to identify the active compounds and test their safety and effectiveness. But it shows how plants can enter the early stages of cancer-drug research.
Why Forest Loss Matters To This Research
Deforestation can remove species before researchers have a chance to study them. The Zero Carbon Analytics analysis estimates that continued deforestation could put USD 174 billion in potential pharmaceutical value at risk by 2050. Stopping deforestation by 2030 could preserve about USD 79 billion of that potential value.
These figures are about possible drug discoveries, not confirmed cancer treatments. For cancer research, the concern is simpler: losing forests could mean losing plants and compounds that might lead to new anticancer drugs.
Technology can help scientists study natural compounds, but it cannot replace species or compounds that were never studied. Natural products are only one part of cancer-drug research, but the history of vincristine, vinblastine, paclitaxel and camptothecin shows that they have already contributed to cancer treatment. The question is how much potential could be lost before researchers get the chance to study it.
Health
Deforestation Could Erase the Biological Library Behind Future Medicines
When we think of deforestation, we often think of lost habitats, disappearing wildlife and declining biodiversity. For the pharmaceutical industry, a forest is a vast biological library. Plants, fungi, microbes and animals produce chemical compounds to survive, some of which can be useful in treating human disease. Scientists can use these natural compounds directly as medicines or as starting points for developing new drugs.
This makes biodiversity an important resource for pharmaceutical research, particularly in tropical forests. According to a new report by Zero Carbon Analytics, 56 of the 579 new drugs approved globally between 2014 and 2024 were natural products or close derivatives.
Yet scientists have explored only a fraction of nature’s chemical diversity. Around 99% of plant compounds are still estimated to be undiscovered, leaving tropical forests as a vast, largely untapped source of potential medicines.
Forests are a Natural Drug-Discovery Library
Plants produce chemical compounds to defend themselves against pests, disease and harsh environments. Some of these compounds can also have medicinal properties.
Vincristine, for example, comes from the Madagascar periwinkle and remains on the WHO List of Essential Medicines. It has been widely used to treat childhood leukaemia, with no full replacement yet available.

This means forests are valuable for the species scientists have yet to study. Every species could hold compounds that become important to future medicine.
What Happens When Forests Disappear?
When forests are cleared, scientists can lose access to plants and other organisms before their potential is ever studied. The chemical and genetic information they contain can disappear with their habitat. Researchers describe this as losing an “option value” — the opportunity to discover and develop medicines in the future. Once a forest is gone, that opportunity may disappear with it.
This does not mean every species lost would have produced a major drug. The study measures potential future discoveries, not guaranteed medicines. But that uncertainty is precisely the problem: scientists cannot know what they have lost if they never get the chance to study it. The new analysis estimates that tropical forests could contain around $714 billion in potential commercial pharmaceutical value from undiscovered drugs in flowering plants alone. Their broader social value could be around $7 trillion.
Deforestation is already reducing this potential. Since 2001, forest loss is estimated to have put about $86 billion in potential drug-discovery value at risk.
The Risk Extends to Future Medicines
The threat does not stop with what has already been lost. Around 13% of the world’s tropical forests have been affected by forest loss since 2001, according to satellite data cited in the study. If deforestation continues at current rates, another $88 billion in potential commercial pharmaceutical value could be put at risk by 2050. The cumulative value at risk since 2001 could reach $174 billion, with the upper estimate reaching $287 billion.
Halting deforestation could preserve much of this potential. If forest loss were stopped by 2030, the researchers estimate that $79 billion in commercial pharmaceutical value could be preserved by 2050 compared with business as usual.
The social stakes are even larger. Under business as usual, another $880 billion in social value from potential medicines could be lost by 2050.
Technology Cannot Replace a Forest
AI, genomics and digital sequence information are helping scientists search nature’s biological library faster. Genetic data can reveal promising biological pathways, while AI can help researchers analyse huge genomic and chemical databases. But these technologies still depend on biological information originating in nature.
Once a species is lost before its genetic information is recorded, technology cannot recover it. This also raises equity questions: tropical forests are in the Global South, while most pharmaceutical profits flow to companies in wealthier countries. Indigenous communities have long contributed knowledge of medicinal plants but receive little of the financial returns.
Protecting forests, therefore, is not only about conserving biodiversity. It could also mean protecting the biological foundation of future drug discovery.
-
Math1 week agoThe 2026 Fields Medals: Four Proofs, Four Decades-Old Problems Solved
-
Society4 weeks agoWest Asia Crisis: Can Kerala’s Returning Gulf Migrants Find a Future in the Green Economy?
-
Space & Physics2 months agoIndia Semiconductor Mission: ‘It’s Not About Fabs. It’s About Building An Entire Ecosystem’
-
Climate2 months agoThe Climate World Cup? How Climate Change Could Affect Player Performance at the 2026 World Cup
-
Society3 weeks agoWhat Is Civilisational Diplomacy? Understanding India’s Newest Foreign Policy Tool
-
Society2 months agoFrom Bell Labs to the Classroom: A Second Career in Teaching
-
Space & Physics2 months agoEngineers Develop Dual-Mode Propulsion System for Next-Generation Small Satellites
-
Sustainability4 weeks agoSharing Over Shopping: How Kerala’s Swap House Is Modelling a Different Way to Consume


