Post by Erik David Johnson
When data-driven, intelligent solutions need to be able to do more than streamline
2019 January 12th
In recent years, more and more innovative ways to make good use of one’s data have been developed—not least through machine learning and artificial intelligence.
This development is part of a global trend that puts the Danish model under pressure, because we can’t simply do as the US or China, when we have to morally distinguish between what we can do and what we should do with data.
Data ethics will therefore be an important part of future Danish IT solutions, but what does this increased focus on data ethics mean to you?
Denmark is in a good position when it comes to innovation, digitalization, research, etc., but we are also doing particularly well when it comes to trust. Trust between the citizen and the public sector, between organizations and their members, and between companies and their customers.
Data ethics goes beyond GDPR and other legislation and not only ensures that you as a business stay on the right side of current legislation but also that you maintain the trust of your customers. Confidence that the data submitted is treated conscientiously and with respect, and confidence that this is for the common good.
If you first lose the trust of your customers, the battle to rebuild that good customer relationship is long and arduous. Data ethics is therefore instrumental in your digital business’ customer foundation. This is already the case today, and it will only continue to become more important in the future.
Denmark strives to be a leader in data ethics
On March 12, 2018, the government put together a think tank on data ethics, which was to make recommendations that could enable Denmark to use data ethics as a competitive factor, with Denmark as the guiding star for both Europe and the rest of the world. I myself was appointed to participate as an AI and data specialist, and on November 22, 2018, we were ready with 9 recommendations to the government aimed specifically at Danish companies.
The work of the think tank confirmed to me that data ethics is not a simple area to move into and perform in practice. A focus for the workshops, debates, and discussions we had in the group was also on how to help Danish companies get started in making responsible, data ethical digital solutions.
This is especially crucial for companies that work with, or have a desire to work with, more intelligent solutions. These use customers’ data as the foundation for building complex models that can find patterns, qualify large amounts of data, work with the Danish language, and make predictions of special business value—all disciplines we already work with in Delegate.
Data ethics in practice
In the think that, I was keen to offer all those who daily work in coding and spent their workdays developing these types of solutions, some very concrete tools to make data ethics design considerations.
This ultimately turned into the third recommendation, i.e. to have a “Dynamic toolbox for data ethics.”
In short, those who code the solutions must be able to use this toolbox directly in their work, so that data ethics are not just polished guidelines that are distributed top-down by management. The content and format of the dynamic toolbox will be organized over the coming time, but I can already identify two areas that we use and relate to in Delegate: Transparency and Bias.
When you make intelligent solutions, like the ones mentioned here, you typically build and train a model on data. Not all models are born equal in terms of how good they are at explaining how they have arrived at their results—i.e. how transparent they are.
In practice, data ethics considerations regarding transparency should be included as a sub-criterion for choosing technology.
Bias then describes the situation where the developers’ own assumptions, prejudices, and worldviews creep into the model, which from there can come with unethical results. For example, this could be an AI model that primarily recommends hiring men in an intelligent HR solution, or a robot that evaluates loan applications that develop a racial preference.
Here, those who work with the data foundation can advantageously be educated in the worst bias-related pitfalls.
Data ethics encourages legislation
Even if you fail to see the value in being a data ethics guiding star among your competitors and prefer to orientate yourself more towards the current legislation—it may still be a good idea to start thinking about data ethics.
At all times, ethical discussions have lent the fuel to the engine of legislation. If we look back in history, legislation on slavery, women’s suffrage, gender equality, and many others has been debated in the context of ethics, before it eventually translated into the legislation that characterizes our modern society today.
The same will apply to data ethics.
Our use of data and the intelligent solutions that facilitate it are in such rapid development that legislation has a hard time keeping pace. We can expect that in the future there will be legislation that will change how we work with data. Therefore, a focus on data ethics solutions is also a strategy for securing the future of your current business and data base.