Written by Mikkel Lehmann Nielsen
Published on LinkedIn March 12, 2021
This article is based on an original publication from itb.
In a recent project, my colleagues from Valcon and 2021.AI developed and implemented Artificial Intelligence (AI) aimed at increasing quality and efficiency in Child Abuse and Neglect cases in Norrtälje municipality in Sweden. This article is a brief summary of the case, and why I think this is awesome.
Cases were stacking up
The +66K citizen-big municipality had seen a 66% increase in Child Abuse and Neglect cases since 2014, as well as an increase in case complexity. Cases were stacking up, ressources were limited and the cases were not being solved in due time. Not only did this cost the municipality unreasonable amount of money due to low efficiency, it significantly hurt the children at risk, as cases like these are very time-sensitive.
Artificial Intelligence to the rescue
The solution was to develop a Natural Language Processing (NLP) model, a type of Artificial Intelligence (AI) capable of reading and ‘understanding’ written text. It works as follows:
When a case handler receives a new case, the AI reads it and instantly finds the five most similar cases. This enables the case handler to quickly learn about previous case-outcomes in similar situations, enabling increased speed at which decisions are made, and the quality of analysis over time. As one employee put it:
“Its like having a thousand virtual assistants helping you with each case”
I like to imagine the case handlers flying around Sweden like Supermen and Superwomen, saving children with the power of an all-knowing AI in their earbuds. It probably does not happen exactly like that in reality, but I definitely think the AI can serve as an upgrade to the case handlers superpowers.
What can we learn from this case?
I encourage you to go over the article yourself (in Danish), but below are the three key takeaways that I wanted to share with you.
This is a key example of a case you can solve only by designing using an ethics first mindset. The single most important criteria for the solution development were that the AI had to be ethically sound. To do this, the team followed the EU’s “Ethics Guidelines for Trustworthy AI”, and had many discussions of how they would ensure the solution would not have any ethical issues. The case is a good example of how AI can be used as an assistant, rather than as an autonomous decision-maker, something that is especially important in sensitive cases like this.
Imagine three levels of autonomy: 1) The AI analyses cases and takes decisions completely autonomously. 2) The AI analyses cases and recommends decisions that are then accepted or rejected by the case handler. 3) The AI assists the case handler, but analysis and decisions are made solely by the case handler. From a financial efficiency standpoint, autonomy level one is the best solution, however, it has a social and ethical risk that is simply not acceptable in this case. The solution implemented in Norrtälje chose autonomy-level three, removing the ethical risk that would otherwise be encountered.
This is an awesome example of how AI can be utilised not just for financial benefit, but for social benefit. It is easy to get stuck in a cost-cutting mindset, thinking AI is simply here to cut cost. Sure, AI can cut cost, and is incredibly good at it too, but it can also be a significant revenue driver, it can increase client satisfaction, employee happiness, be of great social benefit, or play a key role in reducing our environmental impact.
This is a fantastic example of how AI is not just an opportunity for private companies, but a key efficiency and quality enhancement solution for the public. In a survey of 25 public leaders in Denmark, 95% said that AI could reduce the increasing pressure in the public sector, unfortunately, 48% also said that they do not believe AI is being prioritised enough, perhaps due to the fact that 92% of public sector leaders don’t have enough knowledge of the potential of AI.
You do not have to talk about the public for long, before encountering inefficiency frustrations. I like to think of our public systems in the Nordics as being among the most innovative and well-functioning systems in the world, but perhaps they could be even better still?
The best time to invest in AI was yesterday, the second best time is probably now.