Big Data and Ethics

Should we be concerned about how our data is being used?

Tanveer Muntaseer 301620006

Simon Fraser University

CMPT105W

Cristina Eftenaru

6/10/2024
 Description

The growth of AI and the emergence of big data have a direct bearing on applied sciences such
as computer science and data analytics. Large data sets are interpreted by these technologies, which
underpin everything from social media algorithms to medical diagnostics. But they have a big impact on
society, especially on consumers, regulators, and tech corporations. For example, a business that
gathers personal information through mobile apps has the power to alter voter behaviour, impact
elections, and even spark privacy concerns that change social norms and legal frameworks. This research
focuses on the commercial sector, mainly IT corporations dealing in big data and AI, as their practices
influence millions of consumers globally. Given their enormous influence over public and private life, it is
essential to comprehend how these firms handle ethical issues like civil rights, freedom of speech, and
privacy. In today's world, when tech companies, programs, and social media are constantly gathering
user data, the ethical issues brought by AI and big data are crucial. Businesses are under pressure to
maintain a balance between innovation and ethics, particularly about consent and privacy, as public
awareness rises. Businesses frequently utilize ethics to improve their reputation, sometimes in reaction
to public outcry or data breaches. People are beginning to wonder whether businesses are using their
personal information ethically and how much of it is being shared as public awareness grows. Businesses
are under pressure to create clearer ethical standards and practices as a result of this increased
awareness. Users, for example, want to know if their data is sold to unaffiliated third parties or used
against their consent. Businesses have a dilemma when they attempt to advance innovation while
maintaining moral principles like user permission and privacy, such as by leveraging data to enhance AI
technologies. I chose this issue because it emphasizes the conflict between rapid innovation and its
societal consequences, as well as the trade-offs between the advantages of technology and ethical
considerations. AI and big data also led me to wonder how businesses defend their usage of it. What
separates "commercial ethics" practised by IT businesses from academic ethics? And what effect do
these moral guidelines have on consumer trust and government policy? Although more businesses are
implementing ethical policies, I believe that there will still be an absence between their declared
intentions and their actual behaviour. My goal is to learn more about how these principles affect
consumer confidence, public policy, and the general function of data and AI in society.

What is Data ethics?

Data Ethics

Commercial perspectives on data ethics are defensive, focusing on a technologically determinist

framing where innovation is axiomatically good and the economic value of data must be realized. Big

tech and advisory firms focus on ethics as a way to build, maintain, or resurrect 'consumer trust', which

is achieved through investment in ethics centres and research within academia. However, without
strong regulation of the technology sector to create trustworthiness, it may be premature to focus on

evoking trust in data technologies.

Data ethics can be seen as a network of nodes representing various streams of thought and

practice. The philosophical node studies and evaluates moral problems related to data, algorithms, and

corresponding practices to formulate and support morally good solutions. The applied ethics conducted

by philosophers, computers, and social scientists focus on how people access, analyse, and manage data

in particular.

Data ethics is connected to civil society advocacy, providing a framework for guidelines to

advance data developments 'for good' across various contexts. In the UK, the government established a

'Council of Data Ethics' in 2016 to address the 'big data dilemma'. The industry node dominates the

network, incorporating advisory services, tech corporations' operations, ethical review and reflection,

and work by specialists to shape corporate processes.

Perhaps the most recognizable narrative for this agenda is articulated by Hasselbalch and

Tranberg, who frame data ethics as a new evolution of the corporate social responsibility agenda,

forming a new competitive advantage':

“A company's degree of "data ethics awareness" is not only crucial for survival in a market

where consumers progressively set the bar, but it's also necessary for society as a whole. It plays

a similar role as a company's environmental conscience – essential for company survival, but

also the planet's welfare”. (Hasselbalch and Tranberg (p 8))

There is a general belief that data security is harder to achieve now than it was in the past. 70% of adults

believe their data is less secure, compared to more secure, or about the same as it was five years ago

when asked. Merely 6% of respondents state they think data security has improved over time.

However, many Americans admit that they are not always vigilant about paying attention to the privacy

policies and terms of service they routinely see, despite the public's expressed concerns about many

aspects of their digital privacy. While 97% of Americans claim to have ever been asked to accept privacy

policies, only roughly 15% of all individuals indicate they read a company's privacy policy before

consenting to it. While 38% of adults claim to occasionally read such rules, 36% claim they never do so

before consenting to a company's privacy policy.

Research ethics

Research ethics are lagging in terms of privacy and ethical standards. Social media platforms like

Twitter and Facebook are often discussed in terms of privacy, but non-personal information can reveal

sensitive information about specific groups. This raises concerns about group privacy and the ethical
implications of research findings that reveal uncomfortable information about groups. Additionally, the

lack of informed consent is a significant issue, as researchers may collect data from social media without

considering it as a breach of research ethics.

Propensity

The movie Minority Report depicted a future where predictive policing could lead to

incarceration without an act committed. However, in cities like Los Angeles, Big Data analytics have

already been used to identify areas, gangs, or individuals more likely to commit crimes, allowing for

extra surveillance. This raises political concerns, as the high probability of a person committing a crime

can lead to public criticism. Similarly, predicting a person's likelihood of domestic violence could lead to

social welfare organizations implementing interventions, but it could also cause stigma and raise ethical

questions about the role of data scientists and intervention thresholds. Big data makes random

connectedness extremely likely, making it crucial to critically observe the potential risks associated with

predictive policing techniques like RIOT.

Evidence of Learning in Computer-Mediated Learning Environments

Evidence of learning in computer-mediated learning environments can be gathered from a

variety of ways. Three main categories will be used to group these sources, each of which includes

information on writing instruction and learning across a variety of subject areas. Table 1 provides a

summary of these.

The technology and data types that fall under each of these main categories of educational data

vary greatly. These types of data can be produced in computer-mediated learning environments and

have been for some time. The ability to generate large amounts of data, do it continuously, in a variety
of ways, and analyse data sets that have been combined and integrated from many sources are all novel

concepts.

Machine Assessment

Over the past few decades, traditional assessments have been transformed by computerization

in two major areas: computer adaptive testing (CAT) for select response tests and natural language

processing for supply response tests. CAT extends long-standing item response theory, where correct

student response to test items varies according to what the student knows or understands (a latent

cognitive trait) and the relative difficulty of the item. These tests provide more accurately calibrated

scores for students across a broader range of capacities, reach an accurate score faster, and are harder

to game because no two students end up taking quite the same test.

In the domain of literacy, CAT and CDT assessments are most frequently used for reading

comprehension, often becoming a proxy for literacy in general at the expense of writing assessments.

These testing processes and technologies are increasingly embedded into pedagogical practice, such as

end-of-chapter tests in e-textbooks, comprehension tests in online reading programs, or quizzes

delivered through learning management systems. A new species of responsive items offers students

immediate feedback on questions, serving a formative assessment function. Machine learning

designed items with well-tested ones to determine their difficulty.

Natural language processing technologies can now grade short-answer and essay-length supply

response assessments with reliability equivalent to human grading. This rebalancing of pedagogical

emphasis in literacy from a primarily receptive mode (reading) to its productive mode (writing) aligns

with "21st-century skills" of active engagement, participatory citizenship, and innovative creativity.

Structured, Embedded Data

Computer-mediated learning can support the generation of innovative forms of structured data

designed into the instructional sequence. Three kinds of technology and pedagogical processes are

discussed: procedure-defined machine response, organization and collation of machine-mediated,

argument-defined response by humans, and distributed machine learning. Procedure-defined processes

are well suited for highly structured domains where evidence of learning is found in correct and

incorrect answers. They include intelligent tutors, learning games, and simulations, which are cognitive

models that lay out the elements of a target domain, anticipating a range of learning paths.

Argument-defined processes involve nonformal reasoning that allows scope for a range of more

or less plausible conclusions. These processes are necessary for complex and contextually dependent

matters that are potentially disputable, involving human judgment and requiring a person to make their

reasoning explicit while at the same time demonstrating awareness of other plausible reasoning.

Examples of disciplinary practice at the school level include scientific or historical arguments, opinion

texts, information or explanations, and narratives.

which is a systematized judgment process that involves anonymous writers and reviewers. Computer-

mediated review processes manage social complexity, including multiple reviews, multiple reviewer

roles, multiple review criteria, quantitative rating plus qualitative rating, and tracking progress via

version histories.

Formative and summative assessment can be achieved through formative and summative

assessment, with each data point being legible for formative and summative assessment. This allows for

a better understanding of individual learner progress and comparisons across groups of students of

various sizes.

Unstructured, Incidental Data

Technology-mediated learning environments, such as learning management systems, games,

discussion boards, and peer-reviewed writing spaces, generate large amounts of "data exhaust" that can

be captured and recorded in log files. This data is unstructured, meaning it is not framed in terms of a

predetermined data model and each data point does not have an immediately obvious meaning. To be

meaningful, the computer's statistical pattern recognition must be trained by human inference. In

empirical practice, structured and unstructured data are generated simultaneously, and much data

might be classified as semi-structured.

Incidental data exhaust can be mined for patterns of activity that predict learning outcomes,

such as drafting, peer interaction, and revision. These data can be used to provide warnings, social

interaction analyses, and retrospective assessment data. Affective states that impact learning outcomes

may also be detectable in patterns of action and interaction.

capture eye movements, gaze, facial expressions, body posture, gesture, in-class speech, and movement

around the classroom. Human interpretation of these data sets is applied to new data sets, and

unsupervised machine learning clusters computable patterns and suggests that a human-interpretable

label may apply to commonly occurring patterns.

Any of these data sources can provide evidence of learning to write and learning in writing

across various subject areas. However, building a more comprehensive view of big data in education

remains a challenge.

Conclusion

The article provides an overview of the development of big data in education, focusing on the

example of writing. It highlights the continuities with traditional data sources and research

methodologies, as well as emerging potentials in novel data sources and analysis modes. The article

highlights the complexity of data sources offering evidence of learning, not only in literacy but also as a

medium for knowledge representations and complex disciplinary performance across various disciplines.

Big data and education data sciences may offer new windows into the dynamics and outcomes of

learning, but much work remains in the nascent field of education data sciences before the benefits of

computer-mediated learning can be fully realized in educational practice.

It explores the potential impact of Big Data on traditional ethical assumptions regarding

individuality, free will, and power. It suggests that children, adolescents, and adults need to be educated

about the unintended consequences of their digital footprints, and social science research should

consider this gap. In law and politics, political campaign observers and think tank researchers will

become data forensic scientists to investigate digital manipulation of public opinion. Law enforcement

and social services will need to re-conceptualize individual guilt, probability, and crime prevention, and
states will redesign global strategies based on global data and algorithms. The essay concludes that Big

Data has strong effects on assumptions about individual responsibility and power distributions, and

ethicists must continue to discuss how to live in a data-fied world and prevent the abuse of Big Data.
 References

Zwitter, A. (2014). Big Data ethics. Big Data & Society, 1(2), 205395171455925.

https://doi.org/10.1177/2053951714559253

Linnet Taylor and Lina Dencik, Constructing commercial data ethics Technology and Regulation, 2020,

https://doi.org/10.26116/techreg.2020.001

Cope, B., & Kalantzis, M. (2016). Big data comes to school. AERA Open, 2(2), 233285841664190.

Big Data Comes to School: Implications for Learning, Assessment, and Research - Bill Cope, Mary

Kalantzis, 2016 (sagepub.com)

Baker, R. S. J.d., & Siemens, G. (2014). Educational data mining and learning analytics. In K. Sawyer (Ed.),

Cambridge Handbook of the learning sciences (2nd ed., pp. 253–274). New York, NY: Cambridge

University Press.

Technology and Regulation, 2020, jan 10

Technology and Regulation (techreg.org)