BACHELOR Assignment
A framework for ethical ai
Type: Bachelor BIT,CS
Period: TBD
Student: (Unassigned)
If you are interested please contact :
Description:
In the transition to a data-driven society, organizations introduce data-driven algorithms often applying artificial intelligence. For example, fifty percent of the municipalities in the Netherlands use data to a certain extent and 56% of governmental institutions perform research on algorithms. Artificial Intelligence, or rather algorithms in general, is not only a technical issue. For a proper implementation and use of algorithms, they are embedded in a socio-technical system consisting of rules and regulations, an organizational structure, etc. In our experience, the rules and regulations and the organizational structure are especially important for properly addressing the ethical concerns algorithms may have, hence a well-functioning socio-technical system should be well-designed for this purpose.
Illustrative examples can be found among fraud detection algorithms developed by municipalities and other governmental institutions. These algorithms typically scan through individuals' data and create a subset of cases with a high suspicion of fraud. This subset can then be investigated further by a human aiming to confirm and decide whether fraud was indeed committed and to collect evidence to build a case. This is done because municipalities have the duty to monitor the proper and fair use of their societal services, hence to find and punish fraudsters. The damage caused by fraud can be huge; estimated at more than a yearly 1 billion euro of benefits fraud in the Netherlands alone. A further reason for the interest in applying data-driven algorithms for fraud detection is that the available resources for human inspection are limited. Municipalities often have only a small team of people responsible for catching fraudsters, so an algorithm allowing them to focus their attention on a small set of suspicious cases is very attractive.
Although attractive and seen as an important future direction, concerns related to the use of data-driven algorithms may arise. For example, the Dutch government came into trouble because of what is now known as the "Dutch childcare benefits scandal" (Dutch: "Toeslagenaffaire") where, between 2013 and 2019, authorities wrongly accused an estimated 26,000 parents of making fraudulent benefit claims for day-care, requiring them to pay back the allowances they had received in their entirety, which drove many families into severe financial hardship. A parliamentary interrogation committee concluded among other things that "unprecedented injustice" was done and that "affected parents did not receive the protection they deserved" by the system. Another notable example is when the Dutch government introduced the Systeem Risico Indicatie (SyRI) system. This was a social security fraud detection algorithm of the Dutch government. It received an ironic privacy prize for the invasion of privacy of people. There were five reasons for this according to Bits for Freedom: citizens were a suspect in advance, it felt like a violation of their privacy, data might have been used without purpose limitation, it might have been discriminating, and it would be the first step towards a control society. Parliamentary discussions were held about this topic. It became clear that in the design and realisation of the system, too little attention was paid to ethical concerns and that valid points raised by the public were insufficiently addressed.
We take a socio-technical perspective, i.e., view the algorithm embedded in an organization with infrastructure, rules, and procedures as one to-be-designed system. We are developing an ethical framework ourselves. The framework consists of five ethical principles: beneficence, non-maleficence, autonomy, justice, and explicability. The framework can be used during the design for identification of relevant concerns so that they can be addressed early in the design of the entire socio-technical system. In this way, it may contribute to a more robust, complete, and effective introduction of such systems, improve the quality of their designs from the ethical perspective, and avoid injustice and associated uproar.
You can assist us in this effort.
Assignment:
There are multiple possible directions where you can join our effort for defining a helpful ethical AI effort:
- Improve the validation with more use cases
The framework has been validated mainly on the mentioned fraud risk assessment algorithms. You could extend this validation with more use cases especially in other domains. This would require interviewing stakeholders using the ethical AI framework, summarizing the outcomes, reflecting on the ethical AI framework and proposing improvements. - Add a detail level with "recommendations"
The current version of the ethical AI framework has the mentioned 5 ethical principles each detailed with "questions", i.e., concerns to ask stakeholders about. We envisage a further detail level with "recommendations". You could for a subset of the concerns find the main solutions or solution directions for those concerns and propose them for inclusion in the framework. For example, you could focus on fairness concerns and propose a recommendation with main bias detection and prevention methods from literature. These should be understandable by stakeholders who often are not very technical nor knowledgable about AI.