Board Governance and Data Ethics
Data ethics is the integration of ethical thinking with the constraints of data and centres on how to make better decisions that ensures ethical outcomes. Data ethics is a broad topic encapsulating data and ethical decisions, data-based judgement, bias, ethics of collecting and analysing data, and the ethics of automation created by data.
The data we collected in 2010 could paint a low quality, black and white, abstract picture of our world and the data could be used to inform, frame or guide. In 2020, the data collected can describe our world in high-definition colour and is being used to mould, form and shape actions in complicated environments. We are accelerating our collection and use of data. Boards need to get deep and dirty with data and the ethics of judgement based on data, it is complex and often requires new skills.
Boards are biased towards professionals who represent experience in finance, legal, operations, marketing and sales. These “core functions” are supported by professionals with information, innovation, technology and HR expertise. Boards will typically have a deep appreciation of and for the expertise in each of the core functions and will have their own experience in each core area to draw on as a reference point. Members feel comfortable understanding and questioning strategy, ratios, metrics, margins and returns. Often the supporting functions are represented by a single professional and there is little broader expertise amongst the other Board members. When data and ethics enter the room, most members feel uncomfortable. Still, there is a chorus of opinions backed from well known academic institutes, leading consultancies, articles and journals but little practical grounding. This paragraph is an observation on the current status and not a criticism, data, ethics and data ethics are new and this chapter details questions we should be reflecting on and asking, as data is now fundamentally a core function.
Lies, damn lies and statistics is an unattributed quote but helpful in the context of discussing data and ethics. The use of statistics to bolster weak arguments is now followed by the use of data to create the outcome you want. Like statistics which can back any opinion, data can be found to support any recommendation. The most fundamental question a board needs to wrestle with; “How do we know if the decision we are being asked to approve is only based on data that creates that recommendation?” Data is no different to any question a board has been asked to approve previously, however according to Mckinsey 95% of decisions made in business are based on financial data, which represents, according to Google Cloud, 3% of the data available. We accept that there is bias in the data that presented a board paper that came from this financial 3% of data but as the remaining 97% of data becomes available to the board for decision making, it is not going to get any easier to unpick if we are being asked to make the right “ethical” decision.
It is worth noting that data is not easy to understand and in many situations, not understood even by experts and professionals. Data is not oil, sunshine, time, labour or any other of the analogies which all fail. Data is Data. Data is closer to the discovery of a new element or quantum particle which has its own unique characteristics. Data itself creates data, and everything is data. Insight, knowledge and wisdom built on data, just create more data. Decisions are just more data. Data means that bias is amplified, which can hide ethical problems. Our systems are built on discrimination.
Like Data, Ethics is equally not easy to understand and in many situations not understood. However, as we study and gain a better understanding of human behaviour from more data, it becomes increasingly evident that determining an ethical outcome for two people is complicated, for society it is complex, and for everyone it is impossible. We must be able to explain and justify the rationale for any decisions which affect others. However linking our judgement and decision-making to a framework that our ethics or culture can be applied at scale, is a delusional claim. We have to get comfortable that our judgement and decisions, based on any ethical code, hold us to a standard by which we must be prepared to be judged by. If we are deemed to be lacking by professional standard, law or society, Directors must be willing to accept the consequences.
Data in this context is all data. All data embraces data that comes from the market, customers, partners, supply chain and internal data from all systems, as shown in the figure below. Data is collected and stored, analysed, and this creates value, insights and margin, which we observe as an outcome. More data comes from understanding how our actions, which are measured outcomes, effect and affect each of the data collection points. Governance and oversight cover how we manage the entire system and process from the collection of data to understand how our decisions impact outcomes.
Data enables a leadership team to qualify and quantify new risks beyond the obvious of data security, cyber-attacks and poor processes. Data enables leadership to enquire about any and every aspect of the business that can be measured, which means that whilst once we were unable to determine the impact of a decision, a Board can receive the direct feedback from the data that supported a decision, to the outcome and determine if the delta needs to be addressed.
Essentially all ethical and moral thinking has the same goal, and there are several diverse ways to help reach a positive impact by considering what is best for others and seeking mutual benefit. Selecting a model for a specific context increases the chances of a successful outcome, which is a positive impact.
There are numerous books in which moral philosophy and philosophers are categorised; however, there are three core schools: virtue ethics, consequentialist ethics, and deontological or duty-based ethics. Each approach provides a different way to understand ethics and has implications on the decision marking process. If I asked the question “What is the best way to achieve a healthy life?” you can respond with one of three approaches: good nutrition, through exercise, and through spiritual discipline. Each being vital but inadequate by itself. It is bringing these and other strategies together that we can make create better outcomes from a complex decision.
In ethics, no one school answers all the issues, concerns and problems raised by a diverse society; therefore many schools need to be considered to reach an ethical decision.
Do we have a data ontology and/or data dictionary. What standard does it meet?
What confidence do we have in the lineage and provenance of data?
What bias could be in the data, what bias is in the data set, what are the implications?
What data do you get from third parties, what is their model and have you asked q2 and q3 of the sources of external data?
Do you need the data you have, is the data valid, how have you determined that the data is useful?
What data is being used to support human and automated decision making?
Ethics and Data
Which framework (from the table below) are individual leaders using, and are we applying all frameworks equally?
What framework does our company and our values align to and why?
Who’s ethics are we guided by, and how do we know?
Who is accountable for the outcome of our ethical choice?
What level of transparency are we using, and is it working for us?
Algorithms & automation based on data ethics
Do we know the true origin of the algorithms used in our company?
Do we know what data was used to create/ determine the algorithms?
How are we sure that the data set used would apply today?
How are we made aware of the unintended/ unimagined consequences?
What process and procedures do we have in place to test and qualify algorithms and automated decision making, and are these biased to certain re-enforcing conformational outcomes?
Analysis and insights based on data ethics
What tools are we using to create analysis and insights based on data?
Are our tools part of the system or independent of the system?
Who and how are analysis and insights checked?
What happens to analysis and insight that does not align to decisions, strategy, outcomes or rewards?
By what process and who controls it, that determines what analysis and insight are conducted and presented?
Governance and oversight for data ethics
Do we have diverse data ethics skills and experience on this board?
Is data ethics working for us?
Who is us in this context?
Of whom are we asking about data ethics, bias and outcomes?
How to we know if the decision we are being asked to approval is based on data that could only create that recommendation?
What checks and balances do we have in place to understand and determine the filters and direction we are being guided towards?
What delegated authority is there that enables ethical questions to be asked and decided on outside of our visibility, especially in regards to marketing, sales, privacy, terms, model and the use of data?
How are we sure and reassured our decisions based on data have a positive impact on our customers, partners and society?
Are we will to be judged by a professional standard, law or society and accept the consequences?
It is critical that we can explain and justify the rationale for decisions which affect others made with data. We must be comfortable that our judgement and decisions, based on our chosen ethical code, holds us to a standard by which we must be prepared to be judged.
If we are thought to be lacking by professional standard, law or society, we must be willing to accept the consequences.
If you wish to read wider on Data Ethics below are a few books from over 50 on this topic I have read. If you only want to read one, pick “Invisible Women: Exposing Data Bias in a World Designed for Men” By Caroline Criado Perez.
Kearns and Roth “The Ethical Algorithm”
Dignum “Responsible AI”
Hannah Fry “Hello World: How to be Human in the Age of the Machine”
Bartneck,Lütge, Wagner, Welsh “An Introduction to Ethics in Robotics and AI”
Cathy O'Neil “Weapons of Math Destruction”