What this is about:
A University of Cincinnati explainer makes the case for neuro-symbolic AI: systems that combine neural networks’ pattern-spotting with symbolic AI’s rules, facts and logic. Or, as the piece puts it: “neuro” learns from data; “symbolic” deduces from what it knows.

Why it matters:
This is the technical pivot behind a lot of current AI safety optimism. Standard LLMs are powerful pattern machines, but they are also statistical, opaque and prone to confident nonsense. Neuro-symbolic AI promises something clinicians actually care about: not just an answer, but a route to the answer — rules, constraints, traceable reasoning and, ideally, fewer hallucinations.

What’s happening:
University of Cincinnati frames neuro-symbolic AI as especially relevant for healthcare, law, robotics and scientific discovery — all domains where “sounds plausible” is not good enough. In health, the idea is clean: let the neural layer process messy inputs such as text, scans or charts, then let the symbolic layer check outputs against medical knowledge, guidelines, ontologies or rules.

The evidence:
There is some substance here. A 2025 Communications Medicine paper tested a neuro-symbolic system on 206 prostate cancer imaging reports and found it outperformed GPT-4 alone, matched or outperformed physicians on structuring and analysing reports, produced an auditable reasoning chain, and blocked reports with residual identifiers from being transferred. That is exactly the sort of narrow, high-stakes use case where this architecture makes sense.

The catch:
Explainable does not automatically mean correct.

A symbolic layer is only as good as the rules, ontologies and assumptions inside it. If those are outdated, biased, incomplete or badly fitted to the patient in front of you, the system can become wrong in a more respectable outfit.

Reviews of neuro-symbolic AI still flag gaps around benchmarking, uncertainty, trustworthiness and intervenability; XAI research in clinical decision support also shows that explanations often fail when they do not fit real clinical workflows.

The bigger picture:
AI in health is moving from “chatbot answer” to clinical infrastructure: documentation, triage, research extraction, image interpretation, decision support.

The DECIDE-AI and CONSORT-AI guidelines both stress that clinical AI needs proper reporting, human–AI interaction analysis, error analysis, workflow context and real-world evaluation — not just a clever demo.

Reality check:
Neuro-symbolic AI may be safer than a free-floating model in some settings. It may reduce certain hallucinations, improve auditability and make it easier for humans to intervene. But it does not remove the old clinical questions: who built the rules, who updates them, whose bodies are represented in the data, what happens when guidance conflicts, and who is accountable when the system is wrong?

WHO guidance on AI in health keeps returning to exactly this: safety, governance, equity and human responsibility do not disappear because the model got more sophisticated.

The interesting bit:
The seductive promise is that neuro-symbolic AI can make machines reason more like professionals. The quieter risk is that it makes machine reasoning feel professionally legitimate before the evidence has caught up.

The line:
A model that can show its working still needs someone qualified to mark it.

Sources: University of Cincinnati | Communications Medicine | Springer review on neuro-symbolic AI | Neuro-symbolic LLM clinical medicine review | XAI clinical decision-support review | DECIDE-AI | CONSORT-AI | WHO guidance on AI in health

Do you love Therapist Brief? So do we. And we want to keep growing so that we can offer more content and perks for this community—but we need your help.

So if you have a friend who identifies as a fellow salty therapist, or a colleague who is always striving to keep up with emerging research and trends in our profession, consider sending them our way. 

When you share Therapist Brief, you’re connecting your people with a community of thoughtful therapists who take the work seriously (but not themselves). You’re also helping us grow the newsletter so we can keep delivering depth & dopamine in your inbox weekly. 

To say thanks, we’re holding a subscriber drive book raffle. One lucky winner will receive a $30 gift card to their favorite independent bookstore.

Entering is easy.

All you have to do is send your friends and colleagues your unique referral link. For each one who subscribes, you’ll earn 1 entry into our giveaway.

Important note: we can only count your referral once your friend confirms their subscription. Make sure they check their email and complete this important step!

The winner will be announced on July 22nd.

You can also copy & paste this link to others: {{rp_refer_url}}

I’m writing this fresh back from Day 8, Manic Monday at Wimbledon with my Dad. Strictly speaking, Manic Monday is not quite the old Manic Monday anymore, now that Wimbledon plays on Middle Sunday, but it still carries that second-week charge of everyone being overexcited and overinvested in grass. Traditionally, it was the day all the men’s and women’s fourth-round singles matches happened at once, which is both glorious and deranged. Very Wimbledon.

We were on Court No.1, and have been absurdly lucky to have had a run of show courts over the last three years. The highlight was Flavio Cobolli vs Alex de Minaur, which may honestly be the best live tennis we have ever seen. The rallies had us out of our seats. It was epic on every level. Madison Keys’ match was a downer; I really felt for her. We were sat behind her and you could feel the irritation coming off her, that particular psychological violence of knowing you can play better and not being able to access it in the moment. Taylor Fritz then mashed Alexander Bublik with aces, so fair play to the US, some great tennis there, requiring no Presidential intervention (wink).

I wish I could bottle the feeling of Wimbledon. It feels like one of the last bastions of goodness in this country. Polite queueing, strawberries, grass, silence before serves (bar my Dad’s phone going off, which was mortifying), collective groaning at a missed forehand. Even the crowd has its own moral code: if one player gets noticeably more cheers, we will absolutely rebalance the scales and start roaring for the other one. It is deeply British and, frankly, quite beautiful. Then you leave and remember the rest of the country is on fire in seventeen different directions.

Speaking of which, today brought fresh hell: a saga even Trump hasn’t cooked up yet. What to do when you’re a Member of Parliament and a parliamentary standards process is looking into you? Option one: allow due process to play out and perhaps say, “Alrighty then, I may have f’d up here.” Option two: resign as an MP, trigger a by-election, then stand again yourself in said by-election, and use the whole circus to rile the country into another anti-establishment fit at great cost to the taxpayer. Of course, Nigel Farage (leader of Reform UK) went with the latter. Yes, this is really happening.

Rebecca and my profanity-strewn conversation during this evening’s swim could probably be heard across the Cotswolds. I’m not sorry. The fact that so much of our media continues to offer this much oxygen to a man who is not even the Leader of the Opposition is galling. Terrifying, actually. The absolute state of it. I never thought I’d find myself even remotely hoping the Tories (the Conservative Party for US friends) might be the least worst mechanism in a political farce, but here we are. Rant over. For Today.

I’m back to working from my NHS office tomorrow, where I will cook in 30+ degrees Celsius heat, because what even is Fahrenheit (sorry not sorry Ann), in a room that is essentially a south-facing greenhouse with no air con. Today we had to make a business case for air con. Yes, US friends, a sodding business case. Not “humans are melting and clinical work requires a brain temperature below fondue,” but a business case. Unreal.

It’s been a bit like whiplash tbh: one day I’m watching a crowd with a moral compass and kindness, instinctively correct itself when support gets too lopsided, and the next I’m back in a country where Farage can turn accountability into theatre and the NHS needs a business case before it agrees not to boil its clinicians alive. Honestly, the strawberries had better governance.

Last weekend, a white supremacist group marched through Washington, D.C. One widely shared image showed a Black woman on the Metro, surrounded by masked members of the group. Another image shows a Filipino man who is also gay, similarly riding the train and looking terrified.

And it was not just one image or one event. For some clients, the weekend may have also included fear about immigration raids, increased ICE activity, or the steady accumulation of headlines that make people feel less safe.

Why it matters for therapists:
Images and events like this do not stay “out there.” They can come home with people: into sleep, concentration, parenting, work, intrusive thoughts, irritability, vigilance, numbness, grief, exhaustion, and the familiar feeling of having to keep moving through a world that does not feel safe, even when the connection is not fully conscious.

For some Black, Asian, and Latine American clients in particular, these may not land as abstract news events. It may land as something historical, communal, personal, and bodily all at once.

The tool:
Bring up what may already be on the client’s mind. Make room without making the client perform.

Try:

Why this helps:
Bringing up current events that may affect clients emotionally and relationally can communicate a few important things:

Notably, the point is not to prove that you are aware, informed, or one of the good ones (if your ego goes there, no shame - just notice it, escort it to the back of the line for later processing, and know this is part of our work).

If they want to talk about it:
Stay close to their experience with curiosity and openness.

If they do not want to talk about it:
Believe them and move on.

The clinical move:
In simplest terms: open the door, don’t drag them through it.

Remember:

A new BMJ study looked at what happened after Australia’s Social Media Minimum Age Act went into effect. The early answer: not much changed, at least not yet. Researchers found little evidence of immediate reductions in social media use among adolescents under 16.

Here are the details:

The nuance:
This does not necessarily mean the policy “failed.” It may mean the policy was never fully delivered.

Public health interventions can fail because the idea is bad. They can also fail because the implementation is weak, incomplete, uneven, or easy to route around.

In this case, the researchers and accompanying BMJ editorial emphasize that widespread circumvention and low compliance make it hard to know whether the policy itself is ineffective or whether the machinery needed to make it work was not really operating.

In other words, passing a law is not the same thing as changing a system.

The landscape:

Why it matters for therapists:
Parents may hear “social media ban” and assume a problem has been handled. Teens may hear “social media ban” and immediately begin workshopping loopholes with the focus of a doctoral candidate.

Often, we’re the bridge that connects these two parties.

For clinicians working with kids, teens, and families, this study is a useful reminder that online safety and the consequences of social media engagement are ongoing conversations and assessment topics, even when laws, parental controls, or platform policies are in place.

The bigger question:
The BMJ editorial points out that future evaluations need to look beyond simple “hours online” measures. If teens leave one platform and move to messaging apps, gaming spaces, private browsers, VPNs, or less regulated platforms, the harm may not disappear. It may just move.

TikTok, Snapchat, YouTube, Reddit, Discord, WhatsApp, gaming chats, and private group messages all create different risks, different social pressures, and different clinical questions.

The clinical takeaway:
For families, the better conversation may be less “Is my child technically allowed to be on this platform?” and more:

Sources: BMJ Study | BMH Editorial

I spent the past week and a half road-tripping through Northern California with my family, which included college tours, a Dodgers win over the A’s in Sacramento, and a trip to a few national parks.

The college tours were so lovely that I briefly considered whether I should go back to school, but my kids said no. I wouldn’t need another degree, necessarily, but who wouldn’t want to take a ceramics class somewhere with built-in coffee shops and beautiful trees?

We also made it to the Redwoods, where I achieved a lifelong dream of visiting Endor, or at least getting close enough to where it was filmed for my inner child to count it. When I was little, I wanted to live among the Ewoks. Then my brother reminded me they eat humans, which does complicate things. The scenery was gorgeous, though.

Yes, my badgers have had babies and I am the most proud mummy to my furry garden dogs. That is what I call them. I will not be taking feedback.

We may now have upwards of ten badgers in circulation. Is Max slightly too scared to remind me of the unholy amount of working dog food we are now buying? yes. Does he perhaps have some concerns about the state of my brain when I am leaning out of the kitchen window throwing apples for them to catch, as though they have not already been adequately catered for? Yes again.

But my God, they are so cute I might actually die.

Stories from the community

The results:

The people have spoken, and the people sound tired.

Community Responses:

“Burnt Toast to be precise!”