-
Notifications
You must be signed in to change notification settings - Fork 0
Ethical Issues
Following here.
Identify any ethical issues raised by your software. For example:
- Are users' privacy expectations met by your software?
- Could the use of your software result in racial, gender, religious, or any other type of discrimination? How does your software try to mitigate this problem?
- Can your software by abused by some users to cause harm to other users? or to the public at large? How do you mitigate it?
Delta strives to keep user data protected and inaccessible while enabling users to achieve their end goals. To achieve this, Delta implements token authorization using Knox. When registering for the first time, users will be given a unique token that will then authenticate them during the login. Upon each additional Login users will receive a unique token that will allow them to authenticate themselves when using the application. When users logout, the token is destroyed. This process of generating unique tokens to access the app and destroying them when logging out, or perhaps by a timeout or period of idle use, ensure that there is no perpetual one-to-one relationship between a user and token that could be exploited.
Tokens are also stored in a secure hash form. This additional layer of encryption provided through Knox ensures that even if the database is acquired through a security breach, a hacker will not be able to login.
Given this implementation, users can assume that in the event of a data breach and their identifying information being stolen, their data is secure within the Delta App and only accessible to them.
Delta is not responsible for third-party failures, i.e. if the server provider is subject to a data breach, that is out of Delta’s control.
When registering, Delta will require users to adhere to a list of terms and conditions specifying exactly how data entered is used.
As stated above, users can rest assured that passwords and other important login information is securely stored. However, users must be aware that whatever they decide to make public shall be public for anyone else to see. It is important that users know not to post content that may reveal important personal information. This is a responsibility of the user, not of Delta. While Delta allows for users to delete content, the best way to prevent these types of breaches of privacy is to not post any condemning information in the first place.
To log in, Delta requires first name, last name, username, email, and organizational status. If users do not wish to provide this information, they will unfortunately be unable to sign up. However, only the username and organization status will be shown to other users.
When uploading data, users are prompted to give more information about the participants in the studies. The information collected will depend on the study, but common examples would be participant gender, age, weight, or handedness. Participant names will not be stored so as to protect anonymity. When uploading data, you are able to change visibility to public, private, or public to your organization.
Could the use of your software result in racial, gender, religious, or any other type of discrimination? How does your software try to mitigate this problem?
As Delta allows users to interact with the system and produce content that may be visible to other users, there is of course a potential for discriminatory posts. Delta allows admins to delete posts or other discriminatory content, and will in the future implement a reporting feature so that users can report other users if such content is found. Delta will primarily rely on admins or other users in identifying these issues.
- In general, AI models can sometimes be affected by differences in race or gender, as they are based off of the information fed into them. Minority races that have less data in the model could therefore be discriminated against in any difference that could be caused.
- Users can wear smartwatches on their right/left hand and on the outside/inside of their wrist. This differences in orientation affect data collection and could possibly affect our model performances.
- People suffering from disabilities could see data incorrectly skewed or not line up properly if their disability causes what would otherwise be considered abnormal movement patterns in someone not suffering from these disabilities.
Delta's method of mitigating the above listed risks would be to require an inclusion of the salient attributes of the population that the data is taken from within posted datasets. This way, users could use that data to determine if a dataset is right for their use case or not.
Can your software by abused by some users to cause harm to other users? or to the public at large? How do you mitigate it?
As the main goal of Delta is to function as a repository of data for researchers or data scientists, by posting false or otherwise misleading data, a potential for harm to society is present. To moderate this, we will implement a review system of posted data by using a 5-star review system, such as that found on Amazon. This shall serve as a way to prevent spreading of low-quality data. Furthermore, datasets will also be able to be reported by users if certain obviously false or misleading aspects are found.
In addition to giving 5-star ratings, users will be able to comment on posted datasets with the goal of constructive feedback. This opens the door to potential cyber-bullying via offensive and hurtful comments towards other users. To deter cyber-bullying, we will have a report feature for users so their actions can be flagged, with punishment ranging from temporary to permanent bans. Users will also be able to delete comments as they please, hopefully mitigating the lasting effect of negative comments. Organizations will also have the ability to ban/block users from interacting with their community at their own behest. This will allow official organizations, for example say a university, to maintain a respectful environment according to their own rules.