Photo by COGNITVEWORLD
Tunisia has made progress in handling its economic crisis, but it needs bolder policy reforms. According to the International Monetary Fund (IMF), this involves taking major risky fiscal reforms like reducing energy subsidies, public wage restraint and pension reforms. However, Tunisia’s parliament lacks an effective analytical tool to make “good” decisions.
The problem: scarcity of legal information, administrative hurdles and fake information
Parliament has an autonomous and participatory role in making financial governance decisions. This means that it acts as a counterbalance to the government’s decisions.
However, parliament has a huge administrative hurdle. It relies on government reports for information about the budget and policy follow-up. Parliament also does not have its own sources or mechanisms to gather information. This can predispose legislators to error and poor decisions on public finance and public policy.
Even when parliament attempts to tackle the mammoth task, it is subject to constant time pressure. This results in them making decisions that are strongly influenced by heuristics – which are mental shortcuts that lead to “biases”, which leads to sub-optimal decisions.
Furthermore, research has shown that people tend to be more cautious in choices involving sure gains, and risk-seeking in choices involving sure losses. Legislators are no exception.
Solution: Harnessing Artificial Intelligence for timely and salient information’s
BSDL developed an investigative tool that enables parliament to process the government’s information in an efficient manner…gathering timely, accurate and exhaustive evidence, and managing the database. It also uses behavioural science techniques of framing and salience in analysing and presenting the information.
Behavioural Science Design Lab (BSDL), which is a research and advisory firm in Tunisia, tested some solutions that might “de-bias” legislators. Gathering insights from technology and behavioural science, BSDL developed an investigative tool that enables parliament to process the government’s information in an efficient manner.
The prototype uses Artificial Intelligence (AI) to gather timely, accurate and exhaustive evidence, and manage the database. It also uses behavioural science techniques of framing and salience in analysing and presenting the information. These allow parliament to push organisational boundaries, and rethink its oversight model as it will be able to conduct transparent parliamentary debates.
In addition, the prototype tool also reduces concerns about setting aside a budget for hiring parliamentary assistants as data crunchers.
This idea was inspired by firms such as Netflix, ALIBABA, “Accenturazon” and Amazon, that use AI to manage risks and help their managers to make training decisions. AI is able to do this because it supports environments that are fast-changing, and require fast learning.
Results: Innovation for the parliament administration
Our trial raised concerns that need to be addressed when introducing AI into a political process of decision making.
Procedural consideration: Ultimately, the people still want human leaders. Thus, our recommendation of AI as an auxiliary helper instead of as a leader. AI still relies on human oversight committee to decide and approve what gets “autonomized”.
Ethical consideration: The identity of who develops and trains the AI is important, to ensure that the technology is as free from political ideology as humanly possible. We propose that data scientists be the interlocutors and ambassadors between the human oversight committee, and the targets of implementation.
Debiasing AI is an ongoing challenge, especially for some facial recognition tasks that have revealed a strong bias among certain demographics. Reducing pitfalls in information processing tasks, as in our case, could be met by training and the understanding of the human contribution in its oversight, as well as, the interaction AI-Humans.
For political branding purposes, data scientists would be the face of the early outreach strategy, in which parliament would be “co-opting innovation rather than “adopting” the technology.
For Tunisia, Robo-helpers would be better than Robo-gov
Robots offer a learning opportunity to simplify administration and support accountability in the law making life-cycle. In short, efficient decision-making with lesser costs.
This is because AI is able to analyse a large volume of complex data in limited time. It surpasses the human capacity to process information efficiently, and empowers decision makers to make better-informed decisions, in limited time.
Commentators in countries that are very open to Artificial Intelligence (AI), like India, the UK, New Zealand and Japan suggest that robots could make better policy decisions, a concept that they term “robogov”.
However, concerns about biased AI abound in Tunisia. This is because the use of AI could support shifts in organisational power and influence.
In emerging democracies such as Tunisia, the robo-helper – instead of robogov – approach can help. Further iterations of BSDL’s “robo-helper” prototype could help us to identify ways to train both the users and AI, to ensure that decision making is not skewed in favour of a specific brand of politics.