Skip to content

5. A Critical Approach to AI and Algorithms

Jump to bottom
javieraatenas-pixel edited this page Jun 17, 2026 · 3 revisions

Content adapted from

  • Picasso, F., Atenas, J., Havemann, L., & Serbati, A. (2024). Advancing Critical Data and AI Literacies Through Authentic and Real-World Assessment Design Using a Data Justice Approach. Open Praxis, 16(3), 291–310. https://doi.org/10.55982/openpraxis.16.3.667
  • Atenas, J., & Caldwell, N. H. M. (2026). Chapter 9—AI, data and information literacy for the digital student. In A. Phippen & E. Bond (Eds), The Digital Student (pp. 197–235). Elsevier. https://doi.org/10.1016/B978-0-443-34057-4.00008-0
  • Atenas, J., Havemann, L., & Nerantzi, C. (2025). Critical and creative pedagogies for artificial intelligence and data literacy: An epistemic data justice approach for academic practice. Research in Learning Technology, 32. https://doi.org/10.25304/rlt.v32.3296

Introduction

Artificial intelligence (AI), algorithms, and machine learning are increasingly impacting society, raising questions about regulation, fairness, and accountability. AI systems are often presented as neutral, objective, and efficient. However, critical perspectives highlight that:

  • AI systems are built from historical data, which may embed existing biases and inequalities
  • Algorithms reflect the values, assumptions, and priorities of their designers
  • Data infrastructures operate within institutional, economic, and political contexts

Atenas & Caldwell (2026) emphasise that students need to understand how AI mediates knowledge, shaping what is considered valid, visible, and actionable information in digital environments.

Ethical concerns focus on:

According to Whittaker et al., 2018, AI often lacks due process, accountability, community engagement, and auditing, creating power imbalances (Science|Business) and limiting opportunities for participation. Opacity also prevents people affected by automated decisions from challenging outcomes (DataCamp).

It is therefore critical to remove bias and establish legal frameworks to protect individuals and society.

flowchart TD

A[Human Activities & Data Generation] --> B[Data Collection]
B --> C[Data Curation & Cleaning]
C --> D[Model Design & Algorithm Development]
D --> E[Training Data Selection]
E --> F[AI Model Training]

F --> G[Prediction & Output Generation]
G --> H[Decision-Making Systems]
H --> I[Social Outcomes & Impacts]

%% Complexity layers
E --> J[Bias & Historical Inequalities]
J --> F

D --> K[Design Choices & Assumptions]
K --> F

H --> L[Institutional Policies & Governance]
L --> I

I --> M[Feedback into Behaviour]
M --> A

I --> N[Public Trust & Perception]
N --> D

%% Pedagogical / critical layer
I --> O[Critical Reflection & Literacy]
O --> P[Ethical & Responsible Use]
P --> D

O --> Q[Data Justice Interventions]
Q --> H

%% Styling (pastel colours with black fonts)
style A fill:#FDE2E4,color:#000000
style B fill:#E2ECE9,color:#000000
style C fill:#FFF1E6,color:#000000
style D fill:#E0FBFC,color:#000000
style E fill:#F1E3F3,color:#000000
style F fill:#FFE5D9,color:#000000
style G fill:#EDE7B1,color:#000000
style H fill:#D8E2DC,color:#000000
style I fill:#FFE5EC,color:#000000

style J fill:#FFD6A5,color:#000000
style K fill:#E4C1F9,color:#000000
style L fill:#CDEAC0,color:#000000
style M fill:#BDE0FE,color:#000000
style N fill:#FAD2E1,color:#000000

style O fill:#D0F4DE,color:#000000
style P fill:#FFC8DD,color:#000000
style Q fill:#CDB4DB,color:#000000
Loading

Principles of AI Ethics

At a glance, AI systems should benefit individuals, society, and the environment. Key principles include:

AI Ethics Frameworks

Framework Key Principles Source
OECD AI Principles Benefit people and the planet through inclusive growth, sustainability, and well-being; respect human rights, democracy, and diversity; ensure transparency; maintain robustness, safety, and security; ensure accountability https://www.oecd.org/going-digital/ai/principles/
G20 AI Principles Promote inclusive growth and sustainable development; uphold human-centred values and fairness; ensure transparency and explainability; guarantee robustness, security, and safety; ensure accountability https://www.mofa.go.jp/files/000486596.pdf
Australia’s AI Ethics Framework Human-centred values; fairness and inclusivity; privacy protection and security; reliability and safety; transparency and explainability; contestability; accountability https://www.industry.gov.au/data-and-publications/building-australias-artificial-intelligence-capability/ai-ethics-framework
EU Ethics Guidelines for Trustworthy AI Ensure compliance with law; uphold ethical principles; guarantee robustness and safety; prevent harm; promote secure, reliable, and trustworthy AI systems https://ec.europa.eu/futurium/en/node/6945

This framework provides a structured approach to developing, deploying, and governing AI systems ethically and responsibly, ensuring their benefits are maximised while mitigating risks.

A critical approach to AI literacy goes beyond learning how to use tools. It includes the capacity to:

  • Question how AI systems are designed and deployed
  • Identify issues of bias, fairness, and discrimination
  • Analyse the social consequences of algorithmic decision-making
  • Reflect on ethical responsibilities when developing or using AI

AI, Power, and Knowledge

AI systems are not only technological artefacts but also mechanisms of knowledge production. They influence:

  • What information is prioritised or marginalised
  • How individuals and groups are categorised and represented
  • Which perspectives are amplified or silenced

Atenas et al. (2025) highlight the importance of an epistemic data justice approach, which examines how knowledge is constructed through data and AI, and whose knowledge is legitimised or excluded.

This perspective encourages learners to critically consider:

  • Whose knowledge is included in datasets?
  • Whose perspectives are missing or misrepresented?
  • How does AI shape what we know and how we know it?

Bias, Inequality, and Algorithmic Harm

AI systems can reproduce and amplify social inequalities through:

  • Biased training data
  • Simplified or reductive classification systems
  • Lack of diverse representation in development processes

These issues can lead to:

  • Discriminatory outcomes in decision-making
  • Reinforcement of stereotypes
  • Unequal access to opportunities and resources

A critical approach requires recognising that these harms are systemic rather than accidental, and addressing them requires both technical and social interventions.

Pedagogies for Critical AI Literacy

Developing a critical understanding of AI requires pedagogical approaches that are active, reflective, and socially engaged. Picasso et al. (2024) propose the use of:

  • Authentic, real-world tasks that engage learners with real data and AI systems
  • Assessment designs that evaluate ethical reasoning and critical thinking
  • Collaborative inquiry, where learners explore the impacts of AI on society

Atenas et al. (2025) further emphasise critical and creative pedagogies, encouraging learners to:

  • Experiment with AI tools while questioning their assumptions
  • Explore alternative and inclusive data practices
  • Reflect on their own positionality in relation to data and AI

Data justice provides a critical framework for evaluating AI systems by focusing on:

  • Fairness – Are outcomes equitable across groups?
  • Transparency – Are decisions understandable and explainable?
  • Accountability – Who is responsible for harms or errors?
  • Participation – Who has a voice in shaping AI systems?

A critical approach also involves developing ethical awareness and responsible practices. This includes:

  • Recognising the limitations of AI systems
  • Avoiding over-reliance on automated decision-making
  • Ensuring human oversight in high-stakes contexts
  • Engaging with diverse perspectives in AI development

Clone this wiki locally