What is Transparency in AI Ethics?

We have defined Transparency as one of the core Ethical pillars necessary to create your digital AI Twin, therefore we want to discuss and explain what transparency means in the context of AI Ethics, and our practical commitment to ensuring it in all aspects of our work.

Let's try to bring satisfying answers to theses questions: What does transparency really mean? And how does Spheria plan on implementing Transparency?

What is Informed Consent and how is it key to implement Transparency and trust on Spheria ?

Do you remember how most platforms push you their Terms and Conditions, with walls of texts that you are almost forced to agree with, without exactly knowing what's hidden inside or what are the implications? Well, that's the opposite of informed consent. We hate this and we don't want to do the same on Spheria. We put Transparency as a prerequisite for informed consent.

That means, when you agree to use our product, and its future upgrades, you can be guaranteed that you will have all the information you need to make that decision. We do not hide information we don’t think you’ll like, and there are no huge walls of text we know you will never read.

At Spheria, for our user's consent to be legitimate, it must be Voluntary, Informed, and Continuous. These 3 criterias help us set an ethical consent that is also transparent.

Lastly, we are working on implementing Transparency for all features that require informed consent.
The goal will be to build a layer in the User Interface that gives transparent updates or explanations about all new features with your Spheria AI, while explaining clearly the stakes, how the feature maintains your privacy, and how to have control over them.

Explainability, at the center of Transparency

We have already spoken about the right to Explainability briefly in our Fairness Pillar, but Explainability is equally important to Transparency.

Explainability and Transparency sustain each other, like a positive re-inforcement loop.

The first dimension about Explainability touches the behavior and ethical obligations that we, humans and builders of Spheria, need to behave towards users: we are commited to help users, answer their questions about the product, and make sure, at all times, that their consent is voluntary, informed and continuous.

The second dimension about Explainability touches a simple question that every AI company should always be able to answer: “Where does the answer that the AI gave come from?”
We've seen many AI products struggle to give a satisfying answer, because of the black box behind AI models, and mostly because of the unscrupulous crawl and use of people's data without their consent...

That’s not the case when using Spheria. Any answer you receive from an individual's AI has been directly and purposefully entered by the AI owner themself. Any answer you get from a digital AI Twin is therefore explainable, and transparent.

The final dimension about Explainability involves you directly as an AI-owner: because you are in full control of your AI and your Data, and because your AI learns directly from you, you are able to explain any answer that your AI provides and why it answered like that.

Real implementation of Transparency, Explainability, Informed and ethical consent in your digital AI Twin

Example 1: Giving owners full visibility on how visitors interact with your AI Twin:

When you share your AI with your audience, or when visitors come to your AI, it's primordial that you, as the AI owner, have full visitibility on how your AI interacts with visitors. We built the Heartbeat feature with this clear goal in mind, so you have full visitibility on how they interact with your AI, where they look, and how they browse your Virtual Brain.

But to make this ethical WHILE maintaining your privacy, we had to go the hard way: we got rid of all third party cookies and any third party tracking library, to instead build and implement our fully private one. This means no-one else but you can see, in your Heartbeat section, what visitors do while interacting with your AI.

Example 2: Protecting your AI and data against crawlers

The internet is full of bots, some good and some bad. When making your AI public, it remains pretty much undiscoverable until you share it with your audience. Then, the link to your AI becomes noticed by bots (for link previews, or by email clients to verify that links are safe) or social platforms who monitor your activity and try to squeeze even more data out of you (usually against your privacy).

We protect your Data against these bots with multiple layers:

• we've designed your public profile to show very little public information: your first name, and your public picture (if any) - that's it.
• all your specific information can only be access by talking to your AI, and navigating the UI, which prevents standard bots from collecting anything
• we're also actively identifying and blocking bots before they get to your AI
• of course, (repeating what we've already said many times), your data is fully your own, stored in your own separate database, and only you can access it

The other alternative against bots and crawlers is to make your AI private, and only you will have access to it with your access link.

Example 3: What happens when you delete your AI

To guarantee transparency, voluntary and informed consent, we ask that users send a written request to delete their account. Most importantly, we are transparent about the process, the consequences, and the outcome: the deletion is forever with no byte of information left anywhere. By explaining this, and making it a two step process, we are making this transparent, explainable, and informed.

I hope this gives you a good overview of how Transparency plays an essential role in building an ethical AI product. We do put ethics and privacy at the center of our product and culture. Feel free to ask my AI Twin any question..