Ai brings with it the power to gather, evaluate and act upon the vast quantity of data. This coupled with its inherent self-learning capabilities makes Ai one of the most mistrusted platforms in the technology history. As the Ai research leads to increasingly sophisticated artificial intelligence systems and with the advent of advanced Internet-based data mining techniques in recent decades, privacy and social trust have become pertinent social issues.
A platform is socially trusted and trustworthy when the platform has no security or privacy concerns and is designed and recognised as capable of protecting users information.
Here are few tips while designing such a socially trusted Ai product:
Ensure the product addresses privacy and hacking concerns upfront:
This can be achieved by offering data protection services and warranties bundled with the product. Include checks and balances mechanism into the fabric of the system as well automatic encryption of all the data.
Separate Data types
Only the user’s system should have the ability to assemble data into a cohesive picture. Existing brokerage models allow to pay for purchases without providing credit card details to the seller. The product can be purchased and the seller recieves the payment because of the intermediary. The same thinking could be applied to data. Meaning an intelligenet system or application could provide a service to the user by dealing with a data broker, instead of getting access to all user data, to perform an action.
Make AI motivations and actions transparent
Ensure transparency about Ai capabilities of the product and offer multiple ways for users to provide feedback to the system. System should also inform the user when and where and how the feedback will be acted upon.
Be transparent about Data storage
Explain where the data is stored in an accessible and transparent way and explain where the data will go, for how long and who has access to the data and why.
Allow users to have control over data disclosure
As always, the users will approach the Ai product with a certain amount of mistrust. In this phase of product interaction, the system should require users to disclose only the basic information which is the right amount of information required for the system to perform a set of default functions. As the users’ interaction with the product matures, the system should recommend to the user that by disclosing more information, the system can help the users to perform new tasks which will benefit them. At this point also, the control is with the users enabling them to decide not to disclose more information to the system. Once the users reach such a level of comfort and find value in disclosing more information, they will allow the system to have more of their personal information. This control and direct manipulation capabilities should be available with the user at all points to enable them to reverse any personal information access that system may have, at will.