Evaluating AI for Suicide Response

FEATURED RESEARCH

AI Models Rate Responses to Suicidal Thoughts, But Are They Reliable?

Suicide claimed over 48,000 American lives in 2021, up sharply from a decade earlier, making effective mental health support critical. With access to therapy limited, many people turn to AI-powered chatbots like ChatGPT for help.

Researchers from RAND and Harvard Medical School evaluated how accurately three popular AI models, ChatGPT, Claude, and Gemini, assess responses to suicidal ideation using a standard clinical benchmark (SIRI-2).

How They Did:

• Claude performed best, surpassing mental health professionals trained in suicide interventions.

• ChatGPT scored similarly to master’s-level counselors.

• Gemini ranked lowest, comparable to untrained K-12 school staff.

All three models consistently overestimated the appropriateness of responses, with Gemini showing the largest gap from expert evaluations (36% of responses were rated inaccurately).

Why It Matters: AI mental health tools can expand access, especially in rural areas or for those unable to afford therapy. But inaccurate or overly optimistic assessments pose serious risks for vulnerable individuals.

The researchers emphasized AI’s potential, but warned models must be refined to align closely with expert judgment before widespread use in crisis support.

Improvement hinges on ongoing collaboration between AI developers and mental health experts, making tools safer for those in distress.

For more details: Full Article 

Brain Booster

Select the right answer! (See explanation below)

Opinion and Perspectives

DIGITAL TWIN

Digital Twins in Healthcare: Ethical Considerations for Trans Patients

Digital twins are transforming healthcare by creating AI-driven simulations of a patient’s body, providing real-time data for diagnosis and treatment planning.

But researchers José Luis Guerrero Quiñones and Anna Puzio highlight a key gap: digital twins don’t just represent bodies; they shape how people perceive their own identity, with major implications for transgender individuals.

Why It Matters: Trans people often face challenges in healthcare, from misgendering to discomfort with invasive procedures. Current medical tech rarely reflects diverse gender identities, usually defaulting to a cisgender standard.

Digital twins, if thoughtfully designed, could address this gap by accurately representing gender diversity, making healthcare safer and more affirming.

The Identity Challenge: Digital twins produce more than just data—they shape the user's self-perception. For trans people, whose identities may not align with societal gender norms, a poorly designed digital twin risks reinforcing biases or causing harm.

The researchers argue digital twins should accommodate fluid, evolving identities, allowing trans users to explore and affirm their sense of self without being restricted to rigid or binary categories.

Ethical Considerations: For digital twins to truly support trans people, developers must ensure:

• Inclusive data collection: Avoiding bias by incorporating diverse gender identities.

• Agency and autonomy: Allowing users control over how their identities are represented.

• Protection from discrimination: Ensuring privacy and security, as trans health data can be especially sensitive.

Digital twins offer promising possibilities to enhance personalized healthcare for everyone. But first, healthcare providers and tech developers need to actively consider and include trans voices in their design processes.

For more details: Full Article

Top Funded Startups

Exclusive - State of Healthcare AI

MONTHLY REPORT

February's Latest Investment and Industry Insights

AI healthcare startups raised $1.5B in February, but how it’s being spent is just as important.

Most funding is flowing into AI-driven drug discovery and clinical automation, while areas like surgical AI struggle.

Regulatory and geopolitical shifts are shaping the field:

• U.S. deregulation is easing commercialization, but state rules add complexity.

• Europe’s AI Act raises compliance hurdles but unlocks funding.

• China’s open-source AI fuels competition in drug discovery.

• India is investing heavily in healthcare AI.

Will investors stick with proven startups, or will global players shake things up? Here’s my breakdown of February’s top deals and global trends.

Access Our Report: Full Review

Product Pipeline

ALS MANAGEMENT

Cognixion Launches Clinical Trial to Restore Communication for Late-Stage ALS Patients

Cognixion is testing a breakthrough brain-computer interface (BCI) with AI-powered communication software to help ALS patients regain the ability to engage in real-time conversations.

Unlike traditional eye-tracking systems, which become unusable as ALS progresses, Axon-R combines non-invasive BCI, AI, and augmented reality to enable faster, more natural communication.

The clinical trial, partially funded by ALS Accel, will evaluate how well ALS patients adapt to the system and measure improvements in speed and accuracy.

With the potential to restore meaningful interactions for those locked in their bodies, this technology could redefine communication for thousands living with ALS.

For more details: Full Article

Policy and Ethics

AI TRUST

Patients Prefer AI-Written Messages Until They Know AI Wrote Them

A Duke Health study reveals a key ethical challenge in AI-driven patient communication.

Patients tended to prefer AI-generated messages over those written by clinicians, describing them as more detailed and empathetic. However, satisfaction dipped slightly when they were informed AI was involved.

Those told a human had written the message responded more favorably than those told it was AI, even when the content was identical.

Despite this, more than 75% of patients remained satisfied overall.

Researchers emphasize that while AI can ease clinician workload, ethical AI use must prioritize transparency to maintain trust and patient autonomy.

For more details: Full Article

Byte-Sized Break

📢 Three Things AI Did This Week

• Thousands of U.S. schools use AI surveillance to monitor students, but a shocking Seattle Times and AP investigation exposed massive privacy failures, leaving thousands of sensitive student records completely unprotected. [Link]

• French publishers and authors are suing Meta in a Paris court, accusing the company of illegally using copyrighted works to train its AI models, demanding the removal of unauthorized data and compliance with EU copyright laws. [Link]

• AI-generated "doctors" are scamming TikTok users with fake medical advice, using deepfake avatars to promote dubious health claims—experts warn of eerie mouth movements, unnatural pauses, and wildly inconsistent credentials as key red flags. [Link]

Have a Great Weekend!

Trivia Answer: B) Grizzly bear

By March, grizzly bears start waking up from hibernation, especially the males. They emerge groggy, hungry, and ready to replenish the weight they lost over winter. Mother bears with cubs tend to stay in their dens a little longer for protection, but by April, nearly all bears are out and about, looking for food. So, if you’re in bear country, maybe don’t smell like snacks around this time!