BUSI2001 Chap.9 Ethics in Data Analytics
Ethics in Data Analytics
Ethical failures in analytics, from privacy breaches to biased algorithms, can cause discrimination, lost trust and legal penalties. This chapter is built around the PAPA framework of privacy, accuracy, property and accessibility, which the course uses to evaluate data practices.
It works through a hypothetical automated debt calculation in which averaging annual income across fortnights manufactures debts that do not exist, explains algorithmic bias, causal inference errors and survivorship bias, and turns the workshop's blockchain controversies into explicit trade-offs between transparency and privacy, security and accessibility, decentralisation and regulation.
The practice cases ask you to name the principle breached and design a safeguard someone could actually implement. Each case ends with a safeguard, because naming a breach without proposing a remedy leaves the ethical analysis unfinished.
What this chapter covers
- 01
Privacy
- 02
Accuracy
- 03
Property
- 04
Accessibility
- 05
Algorithmic bias
- 06
Causal inference errors
- 07
Survivorship bias
- 08
Blockchain ethics trade-offs
Worked example · free
Placing four data problems under the PAPA principles
- +1(a) breaches privacy: personal data were collected without explicit consent.
- +1(b) breaches accuracy: decisions rest on incorrect data about the person.
- +1(c) breaches property: the firm profits from data whose ownership was never settled.
- +1(d) is an accessibility issue: access to data, and so decision power, is restricted.
Key terms
- Algorithmic bias
- Systematic unfairness in a model's outputs that arises from biased data or design.
- Survivorship bias
- Drawing conclusions from cases that passed some filter while ignoring those that did not.
- Data democratisation
- Making data widely accessible across an organisation or society so more people can use it in decisions.
- Informed consent
- Agreement given with a clear understanding of what data are collected and how they will be used.
- Proxy variable
- An input that stands in for another characteristic, such as postcode standing in for income.
Ethics in Data Analytics FAQ
Why is Cambridge Analytica used as the privacy example?
The Facebook and Cambridge Analytica case shows personal data collected and shared at scale without explicit consent, then repurposed for political targeting. A recommended reading reports that Meta settled the case for 725 million dollars, illustrating that privacy failures carry legal and reputational costs as well as ethical ones.
What did the Robodebt workshop task ask students to do?
Groups of three or four prepared a short presentation explaining how the Robodebt scheme operated, how it challenged or violated each PAPA principle, the key ethical issues including data accuracy and effects on vulnerable people, broader lessons for big data analytics, and best practices that could have prevented its failures.
Is legal data use automatically ethical?
No. The topic argues that ethical analytics goes beyond compliance to building trust, accountability and responsible innovation. A practice can satisfy the law yet still breach a PAPA principle or surprise customers in ways that damage trust, so ethical review should ask what affected people would reasonably expect.
How can analysts reduce algorithmic bias?
Start by testing outcomes across groups, such as approval or selection rates. Remove or constrain variables that act as proxies for protected characteristics, compare model accuracy with and without them, and document the trade-off for decision-makers. Review the training history, because a model learns whatever patterns, fair or unfair, past decisions contain.
Assessment move
Practise applying PAPA to a new case every week, using news stories or the course readings. For each case write four short lines, one per principle, then decide which breach is most serious and propose one safeguard with an owner, such as a privacy officer reviewing each new data use.
Rework the averaging example with different numbers until you can explain in two sentences why an annual total cannot show when income was earned. Keep a list of the three hidden biases with one business example of each, and check your own assignment analyses against them before submitting.
For the blockchain round, prepare a stakeholder table covering businesses, governments, users and regulators, because recommendations that ignore one group usually move the harm onto it. These exercises also supply ethical content for the group industry report. Collect one recent news story about data misuse each fortnight and write a four-line PAPA analysis of it.
By the end of the semester you will have a bank of current examples to draw on in the group report and in your reflections.
Working through Ethics in Data Analytics in BUSI2001? Sia is AskSia’s AI Business Analytics tutor — ask any BUSI2001 Ethics in Data Analytics question and get a clear, step-by-step explanation grounded in how BUSI2001 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.