Saturday, November 1, 2025

Visual 7: Human clinical decision making with machine intelligence MI in the loop: translating and grappling with human consent

 Human clinical decision making with MI , machine intelligence (aka AI) in the loop:


The human layer and Ux interface



  • "Sometimes the smallest things take the most room in your heart." —
  • Winnie the Pooh
  • Above was Winnie the Pooh translating the Chandogya Upanishad:
  • छान्दोग्य उपनिषद् ८.१.३*

    अथ य एषोऽणिमैतदात्म्यमिदं सर्वम्।
    तत् सत्यम्। स आत्मा। तत् त्वम् असि श्वेतकेतो इति।


Smallest events in human lives  are synthesized to create significant events?


In the human context of privacy, can small measures taken daily go a long way to sustain a long term secure health system experience?


How do we deidentify as per HIPAA, the entire data that is captured into our system 2 healthcare data processing ecosystem?

Can missing the smallest things sometimes take up the most room in our workflow?

Are the smallest things, sometimes the smallest pieces in the puzzle, most rewarding in terms of learning and illness outcomes?


Is the work of AI LLMs as just a machine translator in our multilingual workflow small enough?







Consent form: Machine translation provides an added feature to our informed patient consent form that allows a single click translation to any global language!


Let me know if the konkani seems right!

In case it's not we have a manual back up here used routinely for majority of our patients: 


The above is one layer of explainability and raising awareness about patient rights including right to privacy.

Assignment: Get your LLMs to go through the consent forms linked above and check if they are DPDP compliant and if not ask for a better draft of the above consent form to make it DPDP compliant.


Posted by at No comments:

Visual 8: Role of daily events in clinical decision making and role of visual data capture and representation to generate quick human insights and prevent TLDR

Role of daily events in clinical decision making and role of visual data capture and representation to generate quick human insights and prevent TLDR


In a human centered learning ecosystem, with AI in the loop, manual translation is more common?



Above is a layer of manual human to human translation as well as intermittent problems in an otherwise complex patient with comorbidities (will discuss again in the next layer of AI driven analysis)






Again this patient does have comorbidities related to his metabolic syndrome such as heart failure but then intermittent simple human requirements of explainability manifest in his daily sharing through his advocate such as the one here that manifests in his sleep and meta AI helps not just to translate it but also explain it well.

The role of AI driven infographics in explainability:

The role of AI driven infographics in explainability:




Speaker's thoughts: A picture speaks more than a thousand words!

A video can be time consuming though!

Assignment: Ask your LLMs to gather all the patient data from the case report linked above and rearrange it using AI driven removal of exactly dated time stamps and replacement with unidentifiable event timelines comprising labels such as Day 1,n season of year 1,n.





This patient is an example how human simple explainability backed by scientific evidence can provide a new lease of life to a patient of myocardial infarction who travelled the long distance to our college just for that explainability to strengthen his prior trust in us!

Past published work on similar patient: 

LLM textual explanation followed by translation and then text to voice file for the patient's advocate who like most of us also suffers from TLDR:

LLM textual explanation followed by translation and then text to voice file for the patient's advocate who like most of us also suffers from TLDR:





Also demonstrates AI driven support for insulin dose calculation through human learning around carb counting, accounting for insulin correction or sensitivity factor and insulin to carb ratios to decide the total insulin pre meal dose with scientific accuracy.

Posted by at No comments:

Visual 9: The Scientific analytical cutting layer: GIM November 7

The Scientific analytical cutting layer:



What is the sensitivity, specificity of a CT abdomen in a woman with chronic mild intermittent regular pain abdomen and a vague lump in her abdomen?




Are most drug efficacies simply of marginal benefit to patients?


Individual clinical decision making around antibiotic choices anecdote:




Fever chart 

"@⁨Meta AI⁩ Update:
Reviewed the history and it does look like she began with right lower limb cellulitis and then went on to develop heart failure as similar to our ProJR here: @⁨hu1 and then currently she appears to be having nosocomial sepsis and I'm not sure how she grew klebsiella in her blood culture at the day of admission before she was escalated here on piptaz @⁨hu3 please share her deidentified blood culture report.

Unquoted from:


Global clinical decision making around antibiotic choices anecdote:




"It's 3 AM. You're staring at a febrile patient with suspected sepsis. Culture pending. Your hand hovers over the prescription pad. Piperacillin-tazobactam? Meropenem? The voice in your head whispers: "Go broad. Cover everything. Better safe than sorry."

You write for meropenem. Again.

Here's what that voice doesn't tell you, that, in doing so, you've just contributed to a crisis that's killing more people than you might save."


Unquoted above from the link below:

https://www.linkedin.com/pulse/tales-medical-practice-chapter-11-when-antibiotics-stop-kosuru-kknbc


And AI driven decision support for the whole patient:



Above from the static case report journal published version : 



Posted by at No comments:

Visual 10: Explainability, trust and layers of clinical decision making in pre and current AI LLM era

Explainability, trust and layers of clinical decision making in pre and current AI LLM era:

EBM layer: This layer is the one our clinical decision making lab is largely engaged in although the other two layers are no less important.

We have already shared something around those in our previous demos particularly our two video links shared above.

Human layer: This is the most important layer where clinical decision making actually happens at multiple human stakeholder levels:

Below are recent examples of the limits of scientific explainability and it's effect on human trust.


This was human forwarded through WhatsApp and possibly AI generated. So should we call it human generated with AI in the loop or AI generated with human in the loop? 

Well as mentioned before here : https://medicinedepartment.blogspot.com/2025/11/visual-4-what-is-intelligence-gim.html?m=0 all human intelligence is AI!


How much Trust building can one achieve through Human clinical decision making with AI in the loop?




Human mistrust due to persistent uncertainty due to scientifically limited explainability ?


Images of subclinical hypothyroidism patient data:






Human full trust inspite of persistent uncertainty due to scientifically limited explainability 







Can AI act as a guard rail for human mistrust due to lack of communication and explainability?

All the real patient individual demonstrations above take a closer look at individual patient events producing their unique event data trajectories that are perhaps simpler in terms of explainability and interpretability than what may have happened if we tried to inter connect many more individuals with common attributes to predict individual trajectories based on past similar individual trajectories!

Using the language of what we have labeled as "machine layer":

"While random forests often achieve higher accuracy than a single decision tree, they sacrifice the intrinsic interpretability of decision trees. Decision trees are among a fairly small family of machine learning models that are easily interpretable along with linear models, rule-based models, and attention-based models. This interpretability is one of the main advantages of decision trees. It allows developers to confirm that the model has learned realistic information from the data and allows end-users to have trust and confidence in the decisions made by the model.[39][3] For example, following the path that a decision tree takes to make its decision is quite trivial, but following the paths of tens or hundreds of trees is much harder."


Summary of current clinical decision making workflow:


So What? SWOT 


S

trengths: Human centred management, Creativity 


W

eaknesses : User Interface: Asynchronous, academic flatlands 


O

pportunities : Prelude to the symphony of Singularity 


T

hreats: TLDR, DPDP 

Posted by at No comments:

Visual 11: And last but not the least! Machine layers

And last but not the least!


Machine layers:

The machine algorithm will see you now?



Amazon "Help me Decide"!

👆 Quantitative AI driven clinical decision making is currently here?

Is this analogous to clinical decision making:

Key takeaways:


Amazon "Help Me Decide" uses AI to analyze your browsing history (patient's clinical history) and preferences (check out the word preferences in Sackett's classic definition of EBM) to recommend the right product (diagnostic and therapeutic, lab or imaging as well as pharmacological or non pharmacological therapy) for you with just one tap.



The tool helps customers pick the right product, quickly. 

(System 2 decision making fast tracked to system 1 and closer to tech singularity)?


Personalized recommendations include clear explanations of why a product is right for you based on your specific needs and preferences.

Personalized precision medicine with explainability to gain trust!

algorithms? 

Did patients consent to its use? 

Can we trace how a prediction was made, or who’s responsible when it’s wrong?

Unquoted from:

Posted by at No comments:

Visual 12: Synthetic intelligence and singularity? GIM November 7




If AI is dead is the phoenix that has emerged even more scarier?!


Bloom's level 6 phoenix synthetic AI emerging from the wall of Alex:

AI… is dead...!!

It’s just a statistical parrot, rearranging old data, guessing the next word!

Pattern completion wrapped in a shiny interface!

It doesn’t create… it recombines!

And if you’ve noticed all the tools feel the same,
it’s because they are the same?

Same algorithms.
Same limitations.
Same ceiling.

But behind closed doors…
another class of intelligence is emerging?

Not the slow, predictive logic of yesterday’s AI.

Something faster.
Smarter.

Built to operate without human babysitting!

Synthetic Intelligence doesn’t operate on static instruction queries.

It generates autonomous design pathways,

spawning build chains that reconfigure themselves mid-execution?

It doesn’t just produce code…

it integrates logic, interface, and deployment
into a single… self-evolving process?

No delays. No bottlenecks. No waiting for “the next release”!

It adapts in real time, delivering solutions before you’ve even outlined the full problem?!

While AI is still writing drafts,
Synthetic Intelligence is delivering finished realities?


Artificial imitates.
Synthetic… creates.

This is the fork in the road?

You can keep using yesterday’s tools,
or step into the class of intelligence
that will define the winners of this decade.

Unquote:







“Language is needed because we don’t know how to communicate. When we know how to, by and by, language is not needed.”

16/10, 07:07]hu1: How many people would understand the salience of silence..... When we think like western mind that chatters always and we agree that we have to deal with the chattering mind where CBT etc they propound. But theory of silence and understanding of it is the ultimate theory



[16/10, 08:27]hu2: Agree!

This is a very important narrative (albeit non silent) that is a valid counter to the current narrative of Information science, which posits reaching singularity through verbal communication but at heart all of us know that the thinking mind can't reach there as long as it keeps thinking!

Interestingly keeping with the rest of the session content above:

The language of silence begins once the process of decision making aka cision stops! 

Once we reach tech singularity there will be no need for any further improvement? Can we really say there's currently no need for any further improvement in our plurality driven real world?

Till then humans would need to resolve the "curse of dimensionality" that is a product of so called Western reductionism https://userdrivenhealthcare.blogspot.com/2025/08/udlco-crh-reducing-multidimensional.html?m=1 , but while at this current time we are labeling these Western we know that so called Easterns also had good skills in two dimensional thinking aka academic flat lands that have now expanded globally in an exponential manner with the scaling of two dimensional xy mobile screens that almost every individual in the globe holds onto for dear life! 


Posted by at No comments:
Subscribe to: Comments (Atom)