Ethical Perspectives on Personal Data and Automated Decision
Ethical Perspectives on Personal Data and Automated Decision Making Dr Steven Finlay 15/5/2014 Agenda 1. A bit about ethics 2. Ethics, data and decision making 2 A bit about ethics. Definitions 1. Ethics, sometimes known as philosophical ethics, ethical theory, moral theory, and moral philosophy, is a branch of philosophy that involves systematizing, defending and recommending concepts of right and wrong conduct, often addressing disputes of moral diversity. The term comes from the Greek word ethikos from ethos, which means "custom, habit". The superfield within philosophy known as axiology includes both ethics and aesthetics and is unified by each sub-branch's concern with value http://en.wikipedia.org/wiki/Ethics 2. Its about right and wrong. Ethics is. Subjective, personal, unique 3
Common ethical frameworks Ethics Non-Consequentialist Consequentialist Its more about the journey than where you end up The means justify the ends Utilitarianism Greatest good for the greatest number (Jeremy Bentham and John Stuart Mills) Virtues Virtuous modes of behaviour (Aristotle) (Human) rights Right to life, liberty, property, privacy, etc. (Locke and Rawls) Kants ethical theory
Religious Teaching Universality: Ethical is something all rational people would agree with (e.g. the ten commandments) Golden rule Do unto others as you would have done unto you (Do no evil) 4 Ethics in practice All ethical frameworks have their weaknesses 5 A bit about ethics. Relevance in the real world If I follow all laws and regulations, then thats all I need to worry about right? Legal Lots of laws allow unethical actions to occur:
Ethical It is illegal to give alcohol to a child under 5 Another example is tax avoidance: A great example of what we mean when we talk about the of the law as opposed to the letter of the law spirit 6 A bit about ethics. Relevance in the real world It pays to be ethically minded: Organizations adopting ethical policies tend to reap the benefits. Largest ever study of the relationship between ethical performance and financial performance: Losses from reputational damage, resulting from actions that are perceived to be unethical, are particularly severe. Corporate virtue in the form of social and, to a lesser extent, environmental responsibility is rewarding in more ways than one. (Orlitzky et al. 2003)
7 A bit about ethics. Summary There are many ethical perspectives. We all have our own view on the rightness/wrongness of different actions. Ethical theory is all very well, but putting it into practice is difficult. The world is a messy mixed up place. The one thing that can be said to apply across all ethical frameworks: An ethical action is one which the perpetrator can defined in terms of more than self interest. (Finlay 2000). Ethics pays. A well thought out, well implemented ethical corporate policy benefits both organizations and consumers/individuals in the long run. 8 Agenda 1. A bit about ethics 2. Ethics, data and decision making 9 Ethics, data and decision making. Whose data is it anyway? Utilitarian orientated perspective
My data is a resource to be harvested and put to use. Constraints (laws) to prevent specific abuses and misuse of my data. Better data & predictions = better outcomes. Everyone benefits. Kantian/Rights based perspective My data is a part of who and what I am. Its mine! My data should be treated with respect, just as I expect to be treated with respect. I will decide how data about me is used. You have no right to use my data without my
permission. 10 Ethics, data and decision making. Whose data is it anyway? Approach Pros More/better data means better decision making. More get the very best deals (if they warrant it). Utilitarian orientated Social benefits. More data to perspective support national / community initiatives (e.g. medical research and counterterrorism). Best for the economy. Each individual has control over their data and the uses to which it is put. Kantian/Rights based Less social exclusion.. Right change/withdraw perspective permission to use data, including Right to be forgotten.
Cons People less in control of their own destinies. Better predictions does not always equate to increased in well-being. The have-nots have even less. Once the data is out there, its out there for good. Poorer decisions for individuals may result, if data is withheld or otherwise unavailable. Lower economic benefits. Society as a whole may suffer because large scale studies are data limited. (e.g. medical research and counter terrorism). 11 Ethics, data and decision making. Is more data and better prediction always better? More/better data leads to the promise of near perfect predictions in some areas. Is this a good thing? Sometimes:
Identify terrorist subjects with high degree of certainty Predict that a heart attack is very likely in the next 24 hours Long term compatibility on a dating site .. But not always Near perfect insurance claim predictions are no benefit to anyone (except the insurer) Do I want to know, years in advance, when I am likely to die? .. 12 Ethics, data and decision making. Whose data is it anyway? Whats the direction of Travel? EU has taken a rights based approach, and looks like it will continue to do so, via revised Data Protection Legislation approved in March. USA, has to date, followed a more utility based model. Use data for whatever you want, but we will legislate where needed. 13 Ethics, data and decision making. What data to use when?
Age Alcohol consumption Credit history Criminal records Dependents DNA Driving speed Education Gas consumption Gender Grocery purchases at supermarket Income
Last book purchased Live with smoker (Y/N) Marital status Medical history Music currently listening too Race Religion Sexual orientation Smoker (Y/N) Type of car you drive 14 Ethics, data and decision making. 1. Immutability of data? Mutable (Individual can change easily) Immutable (Individual cant change at all) DNA Income
Criminal record Age Education Alcohol consumption Gender Type of car Dependents Race Smoker Religion Live with smoker Sexual orientation Last book purchased
Gas consumption Music currently Listening too Grocery purchases Medical history Marital status Driving speed 15 Ethics, data and decision making. 2. Beneficiary? For whose benefit is a decisions made ? (This is not the same thing as if the individual benefits from the decision) Decision maker Individual / society Benefit payment Treatment for illness Home improvement grants
Child protection Match on dating site Parole Foreclosure Survey selection Product marketing Selection for tax inspection Suspect selection in criminal cases Redundancy selection Credit granting Making job offers Insurance
pricing 16 Ethics, data and decision making: 3. Impact What is the potential impact of decisions on an individuals well being? Low Impact High Impact Home improvement grants Match on dating site Credit granting Survey selection Product marketing Insurance pricing Selection for tax inspection Benefit
payment Foreclosure Making job offers Parole Child protection Redundancy selection Treatment for illness Suspect selection in criminal cases 17 Ethics, data and decision making. Risk in decision making Decision maker High 3. Impact on individual
2. Beneficiary of decision Low Individual Mutable 1. Immutability of data Immutable 18 You need to decide whats most important within your ethical view (i.e. column order). E.G, foreclosure, redundancy, parole Greatest
More legislation Audit & regulatory oversight Public interest Greater manual involvement Simple and explicable models Judgemental overriding Expert Buy-in Understand model weaknesses Constant monitoring Least Less legislation Predictive ability trumps all else Complex black box models Automated model generation
Rapid redevelopment of models Little oversight Impact of decision on individual Beneficiary Immutability of the of data used decision Ethical challenge / risk High Decision maker High Individual High Low Low
Low Decision maker High Individual High Low Low E.G. Marketing type applications Ethics, data and decision making: Alternative perspective Its nothing to do with the data or the decision maker Its how you make the decision thats important Impartial, data driven process = GOOD (Ethical) Biased/judgemental decision = BAD (Unethical) Example: If women more likely to do X or Y than men (or vice versa), then its fine for Gender to feature in a predictive model, if thats what the data is telling us. However, this view is not popular, at least not in the UK or EU.
As evidenced by (fairly) recent decisions on the use of Gender in insurance, despite gender being one of the most predictive data items for all sorts of insurance claim behaviour. 20 In Summary Ethical data use and decision making brings its own rewards An ethical strategy is about more than just following the law. Ethical and legal is where you want to be Some things to consider when formulating an ethical data and decision making policy: The immutability of the data that you use. The impact that your decisions will have on individuals. The beneficiaries of the decisions you make. 21 Bibliography and further reading Boatright, J. (2014) Ethics in Finance (3rd Edition). Wiley Finlay, P. (2000). An introduction to Business and Corporate Strategy. Pearson Education.
Finlay, S. (2014). Predictive Analytics, Data Mining and Big Data. Myths, Misconceptions and methods. Palgrave Macmillan. Orlitzky, M., Schmidt, F. L., Rynes, S. L. (2003). Corporate Social and Financial Performance: A Meta-analysis. Organization Studies, volume 24, number 3, pages 403-441. 22
PLOT: 4 ghosts (Jacob Marley & the 3 Christmas ghosts) come visit him. Each one of them has a specific message/ lesson. NEW SCROOGE: In the last chapter, Scrooge shows he has changed by doing different behavior than Old Scrooge...
Acids An acid is a substance that tastes sour, reacts with metals, and turns blue litmus paper red. Acids react with certain metals to produce hydrogen gas. Not all metals react this way, but magnesium, zinc, and iron do. When...
Misplaced Trust . The primary issue with a perimeter-centric security strategy where countermeasures are deployed at a handful of well-defined ingress/egress points to the network is that it relies on the assumption that everything on the internal network can be...
monitoring ≠ wilderness character. 2008: Keeping It Wild. 5th Quality not recognized for . monitoring . purposes. Undeveloped included cultural resources. A very brief history: + Some of you may remember from our webinars last year, we really started thinking...
Expected life events Unexpected life events Expected life events Unexpected life events Starting school Having children Getting married Moving House Leaving Home Retiring from work Taking exams Puberty Getting divorced Becoming bankrupt Being made redundant The death of a loved...
silver bullet, only . silver buckshot. Why not more Attention and Action on CC? Business action lacking. Resilience of systems/infrastructure related to food production, energy production, transportation... Investor action lacking.
At ~1.2 GPa the new magnetic transition appear at ~35 K probably related with Co-sublattice and Ce-related transition becomes very broad and is shifted to ~14 K. These big changes in magnetism of CeCoSi are most probably related with a...
1824: The Brooklyn fire department consisted of just four engine companies and one hook and ladder company . 1825: June 25th, $1400 allocated for the use of an engine and a house , located selected to be on the south...
Ready to download the document? Go ahead and hit continue!