# Processing with AI
## Exploration : 👩‍⚖️ Ethics of AI

Name: Granger-Veyron Martin
>
>
Subject: 🔬 Detect rare diseases in medical results (blood tests, MRI, radiology, etc.) using Machine Learning
>

>[TOC]

## Design brief

### Bias

If we don't source our dataset with enough rigor, the following bias might appear:

>1. a Virus or Bacteria
>2. b Patient's age
>3. c Patient's comorbidity
>4. d Patient's past ( parents's sicknesses)
>5. e Patient's gender

We will ensure that our model is not biased by:

>1. categorising every sickness
>2. categorising every patient ( age, gender...ect)
>3. Train the model to understand people's history and risks

### Overfitting

We will make sure our model does not overfit by
> Checking the accuracy of our model on a patient that is absolutely not sick
> Meaning that it will be super important to make sure that our model has the "else" input. In otherword the model due to overfitting could tell us that he recognized a sickness in one patient that is actually nothing. In our course, we saw that a guitar could be recognized in picture where there are just lines of colors. Maybe that a overfitting model could find a lung cancer where there is absolutely nothing. We have to be careful!
> We may also verify that due to overfitting, other sicknesses( a simple fiever for exemple) will be seen as deadly rare deseases. That could also be a problem of bias but it is necessary that we make sure that in both ways, a lung cancer is not seen as a fiever and a fiever seen as a lung cancer.

### Misuse

>We have to remind ourselves that our application could be misused by pharmaceutical company to sell more a particular product for exempke
>Moreover this model is quiet complicated as it has to take into account the medical secret that will enable anyone to have access to any kind of information.

### Data leakage

>**🔐 Closed source:**
>It is important to know that if we implement patient's data in our model, our model has to be a Black Box.
>Medical secret is a very important notion and could even be a problem to create our model. As we are not doctors, we will we be however able to obtain every information that we need so our model is not biased or overfitting.It will of course be complicated to obtain those private information.
>We could however discuss with our patients and see if some of them were to agree on a open source dataset.
>In a catastrophic scenario, where all of our training dataset were stolen or recovered from our model, the risk would be very important.
>The different misuses that we have talked about could also come from a open source data set.
>Patient could feel very uncomfortable knowing that their private information are not secret anymore.
> We could in fact be sued by all our patient.

### Hacking
> If someone found a way to "cheat" our model and make it make any prediction that it want instead of the real one, the risk would be huge > We could in fact talk again about the pharmeceutical company that could hack into the model and target a certain number of desease for which they could offer a drug.
> Since we are talking about life threatning deseases, it could be a massive problem if someone were to hack into the model and change the output of the test.

### My opinion

I would like to conclude on my opinion about a project of this size.
I think this model could be a very useful tool for doctor to confirm what they saw on a IRM for exemple but i beleive that in terms of ethics and in term of medical precision that a doctor can be assisted by a technological tool like this one but never replaced. We are dealing here with the life of many people and it is important to always remember this fact.
I could go further by saying that this model is very useful in a "computer vision project" like for exemple checking during an exam that we are dealing with the right student. ( The tool is, I beleive sufficient)
However, the different bias and overfitting could create deadly consequences. The misuses could make company richer without even caring about the health of our patient, and finally i beleive that asking people for those private information are in a way not ethical. This is why we have doctors.