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

Name:
> Josephine BARBETT
>
>Subject:
>
> 💕 Improve dating apps matching algorithms using NLP

>[TOC]
>
>
> 

## Design brief

### Bias

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

>1. To **avoid fake and hacks** profiles 🎭 👿
>Many people are reluctant to use dating apps because they are afraid of finding fake profils. In fact, some pervert people create fake profile on those dating apps in order to try to appeal women who would absoluletly not interested in those people in normal time. Therefore, many people don't want to try this experience and they don't even want to try to sign in dating apps.
>In order to reduce this bias, the dating apps should ask for more verification documents before enabling anyone to sign in. By doing so, it would be way more difficult for anyone to create a fake profile and to make people waste their time and even be in danger.
>
>2. To avoid **promoting only "classic"** profiles 👘 👩🏾‍⚖️ 👨🏼‍✈️
>In order to make everyone to feel comfortable with using the dating apps, it is essential that all kinds of profiles are represented. So that whether you like any trend, wearing high heels, or any kind of hat, you will find the one who'll like it too. Diversity is a huge richeness that dating apps should developp and promote.
>
>3. To avoid the **collection** of matches 💯💯
>Many people are reluctant to use dating apps because they are afraid that their crush actually has plenty of matches and that they would be only a number among others. Dating apps should implement rules limiting the amount of possible matches at one time. Therefore, the users would be able to match only a certain amount of potential crushes and would fully consider them as human rather than as numbers only.

We will ensure that our model is not biased by:

>1. 🌍 Sourcing our data from a dataset that contains many **different profiles to ensure divertisy**. In fact, it is important that all the minorities will be represented in the dating apps to make anyone comfortable with signing in dating apps. Therefore, we will make sure that our dataset has as many diverse profiles as possible.

>
>2. 📝 Making the users active in the design of the dating app. In fact, I think it would be useful for everyone to include anonymous **feedbacks** possibilities in the app. With this new feature, everyone would be able to leave their feedback about the app that is important for them. This would widen the potential app users and would make the app better known among more people.
>
>3. 💸 In order to limit the amount of fake profile, I think it could be interesting to ask for a **little money** to have access to the potential targets' profile. In fact, I really think that most of the people potentially willing to create fake profiles would be discouraged to do so if they were asked to pay even a small fee to access the other profiles. By implementing this measure, more people would be reassured to sign in dating apps and everyone would be more confident in the dating apps.

### Overfitting

> 🧠💡 The principle of the articifial intelligence is that the models learn on their own and they improve themselves thanks to all the data that they face.
Therefore, it is tough to find an inbetween between collecting to much data and having an app that only contains a little information. Therefore, I think it is important to filter the profiles that would be allowed to sign in the app or not. This matches with our idea to ask the users a little fee to register on the app. Therefore, the app would not be busy and overloaded with fake profiles that don't deserve anything.
If the app contains too much useless data, its carbon footprint will also be way higher. Therefore, implementing this new regulation is a good thing on many points of view.

### Misuse

>👿 We have to remind ourselves that our application could be misused by hackers, fake users or even perverses that would want to trap some people. Therefore, we deeply think that it would be important to ask users to pay a little fee to allow them to sign in the dating app.
>We are aware that we have to pay attention to reduce and to limit as much as we can the amount of fake users that could threaten the security of the others.
>In fact, we know that it can happen. We have many examples, especially women who are harassed by perverses on those dating apps and that they don't feel safe and secured to meet unknown people. That is one of the reasons why it is essential to make female users feeling safe and secured while signing in dating apps.

### Data leakage

> 👮🏼‍♀️ 🏡 In a catastrophic scenario, where all of our training dataset were stolen or recovered from our model, the risk would be that all personal data that the users registered on the dating app would be stolen and potentially used by someone with bad intentions. Dating apps users register some personal data like their sexual orientation or any other personal information that no one should be able to access.
> Those kind of events could also lead to bad intentionaly use them in order to blackmail any user and that can have some huge consequences on many points of view.
> Those bad intentionned people could also use the personal data registered by the app users like their addresses or their phone number to robber them or to steel any useful information.
> That is the reason why we ask a little fee for all our users to be able to afford an efficient computer specialist to protect our users' personal data and to make them feel comfortable with registering personal data that would be useful in the path of fiding a new partner.

### 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 that all our users would leave our dating app in favour of another one. Therefore, it is essential for us to protect our code to avoid anyone to steel it.
> More than the code itself, a robber could stole our code but also all personal data that our users registered on the app. This point can be critical because the users trust us and trust the confidentiality of our app? They do not want their personal information to be stolen by any unknown robber.
>
> Again, this hacking possibility could be avoided by the fact that we ask a little fee from our users. With this collected money, we are able to buy the prestation of a specialist un computing sciences that could ensure that our data is safe and than no one could potentially steel it.
>
>

### References

https://uxdesign.cc/how-to-mitigate-social-bias-in-dating-apps-bcecee369d30

https://www.forbes.fr/business/des-applications-de-rencontre-suppriment-les-criteres-dethnicite/

https://www.scmp.com/lifestyle/family-relationships/article/2168914/dating-apps-can-do-more-fight-racial-bias-and

https://www.axios.com/what-dating-apps-are-doing-to-fight-bias-d0285621-5b5a-4c28-87a7-0881df45d563.html