Table of Contents

Dark Patterns in UX Design.............................................................................................................5

Core Pattern Typology......................................................................................................................5
The Five Dark Strategies..................................................................................................................5
Case Studies & Examples.................................................................................................................5
Data Privacy Laws & Regulations...........................................................................................5
Pakistan's Legal Landscape..............................................................................................................5
International Comparison: GDPR vs. CCPA.................................................................................5
Business ethics canvas..............................................................................................................6
1. Introduction & Context................................................................................................................6
2. Theoretical Foundations...............................................................................................................6
Table 1: Markkula Ethical Framework & Business Analytics Implications...................................................6
Case Study: EuroTravel....................................................................................................................6
3. Business Model Canvas (BMC)....................................................................................................7
4. The Opportunity Canvas (OC).....................................................................................................7
5. The Business Ethics Canvas (BEC)..............................................................................................8
6. Process for Development...............................................................................................................9
Strategic Reflection.........................................................................................................................10
Data governance & The Zachman Framework.....................................................................11
1. Data Governance Frameworks: DAMA DMBOK....................................................................11
2. Peter Aiken’s Pyramid & Four Phases of Data Management.................................................12
Data Governance.............................................................................................................................14
Data Stewardship.............................................................................................................................14
The "Renting an Apartment" Analogy..........................................................................................................14
The 3DC Stewardship Framework...............................................................................................................14
Data Architecture............................................................................................................................14
• Four Architecture Perspectives:.................................................................................................................14
The Zachman Framework..............................................................................................................15
Interrogative Columns (The Questions).......................................................................................................15
Perspective Rows (The Views).....................................................................................................................15
Applying Zachman to Data Ethics.................................................................................................16
The Zachman Framework(summary Nllm)..................................................................................16
Data Modelling.............................................................................................................................................16
• Case Study 1: Gender Representation........................................................................................................16
• Case Study 2: Driver Safety App (Privacy Minimization).........................................................................16
• Case Study 3: Voter Canvassing (Secrecy of the Ballot)...........................................................................17
Data Quality Management............................................................................................................................17
Data Warehousing & Business Intelligence (BI)..........................................................................................17
Data Science and 'Big Data'..........................................................................................................................17
Dama vs Zachman...........................................................................................................................18

1
Ethical issues in AI (case study).............................................................................................19
1. Theoretical Background: Understanding "Future Signals"...................................................20
2. Methodology: The Revised Analysis Framework.....................................................................20
2.1 The Quantitative Foundation (Yoon’s Approach)..................................................................20
2.2 The Qualitative Addition (The Authors' Contribution).........................................................20
Process:.........................................................................................................................................................20
3. Case Study: Ethical Issues in Artificial Intelligence.................................................................21
4. Conclusion and Implications......................................................................................................22
Building Transparency into AI Projects................................................................................22
Core Thesis.......................................................................................................................................22
1. The Problem & A Cautionary Tale............................................................................................22
2. The Four Impacts of Transparency...........................................................................................23
1. Decreases Risk of Error & Misuse: Poor communication between teams leads to failure.....................23
2. Distributes Responsibility: Informed stakeholders can make responsible decisions and be held
accountable...................................................................................................................................................23
3. Enables Internal & External Oversight: Critical for catching ethical, legal, and reputational risks.......23
4. Expresses Respect for People: Preserves human autonomy and prevents manipulation........................23
3. How to Implement Transparency: A Practical Guide..............................................................23
Key Distinction: Transparency vs. Explainability........................................................................24
Conclusion........................................................................................................................................24
Conversations Companies Need to Have":............................................................................24
Core Problem...................................................................................................................................24
Proposed Solution: A Structured, Conversational Framework..................................................24
1. Form the Right Team..................................................................................................................24
2. The Three Crucial Conversations..............................................................................................24
Conversation 1: Define Your Ethical Standards...........................................................................................24
Conversation 2: Identify the Gaps................................................................................................................24
Conversation 3: Understand the Sources & Operationalize.........................................................................25
Summary: "Google Let OpenAI Scrape YouTube Data Because Google Was Doing It Too"
(PCMag)..................................................................................................................................25
Key Points:.......................................................................................................................................25
Core Issue:........................................................................................................................................26
Summary: "Machine Bias" (ProPublica).............................................................................26
Key Findings (from the original report):......................................................................................26
Core Conclusion: The report demonstrated that algorithms can automate and amplify
existing societal biases, posing a grave threat to justice and fairness when deployed without
rigorous auditing for discrimination..............................................................................................26
Overcoming the C-Suite's Distrust of AI...............................................................................26
The Core Problem...........................................................................................................................26
Root Causes of Distrust...................................................................................................................26
Four Actions to Build Executive Confidence................................................................................27
1. Create Reliable AI Models:.....................................................................................................................27

2
 2. Avoid Data Biases:...................................................................................................................................27
3. Ensure Ethical & Moral Decisions:.........................................................................................................27
4. Explain AI Decisions:..............................................................................................................................27
Key Takeaways & Tasks for Leaders.............................................................................................27
Conclusion........................................................................................................................................27
Summary of ethics and AI files..............................................................................................27
Ethics and AI Slides................................................................................................................28
Part 1: Building Transparency into AI Projects...........................................................................28
Introduction:.................................................................................................................................................28
What is Transparency in AI?........................................................................................................................28
Impacts of Being Transparent (The Four Key Benefits):.............................................................................28
Implementing Transparency (What Good Communication Looks Like):....................................................29
Crucial Distinction: Transparency vs. Explainability:..................................................................................29
Part 2: Overcoming the C-Suite's Distrust of AI..........................................................................29
Four Actions to Build Executive Confidence:..............................................................................................29
Key Realities Highlighted:...........................................................................................................................29
Tasks to Raise Executive Confidence:..........................................................................................................29
Part 3: Ethics and AI: 3 Conversations Companies Need to Have.............................................29
Prerequisite: Form the Right Team...............................................................................................................30
The Three Conversations:.............................................................................................................................30
TOGAF....................................................................................................................................30
1. The Four Architecture Domains (BDAT)2.................................................................................30
2. The Architecture Development Method (ADM)10.....................................................................30
3. Key Concepts & Components.....................................................................................................31
4. Why Organizations Use TOGAF...............................................................................................31

.................................................................................................................................................31
5. What’s New in the 10th Edition?...............................................................................................31
TOGAF VS ZACHMAN VS DAMA-DMBOK......................................................................31
Case studies.............................................................................................................................33
Scenario:...........................................................................................................................................33
Business analytics use case 1:.........................................................................................................33
Business analytics use case 2:.........................................................................................................33
Common Themes Across Both Cases.............................................................................................34
Question 1: The Ethics of Trade-offs.............................................................................................34
"In Case 2, the HR manager opts for a weaker open-source security method to save costs. Using the
DAMA-DMBOK framework and the 'Renting an Apartment' analogy, explain why this is an ethical failure
of data stewardship."....................................................................................................................................34
Question 2: Applying the Business Ethics Canvas (BEC)............................................................35
"For Case 1, identify one 'Green' Opportunity and one 'Red' Risk using the Business Ethics Canvas
categories of 'Utility' and 'Rights'."...............................................................................................................35
Question 3: Building Executive Trust............................................................................................35
"The C-Suite is often distrustful of AI. Based on the HBR article 'Overcoming the C-Suite's Distrust of
AI,' identify two root causes of distrust present in Case 1 and suggest how to build confidence.".............35

3
 Question 4: Strategic Alignment (Zachman Framework)...........................................................35
"How can the Zachman Framework be used to address the 'Legal Liability' weak signal in Case 2?".......35
Question 5: Ethical Governance and "The Three Conversations"............................................36
Q: In Case 1, the cybersecurity team highlights the importance of protecting data. Which of the "Three
Crucial Conversations" must the senior team have before deploying this AI?............................................36
Question 1: The Foundations of Data Management............................................................36
Q: Using Aiken’s Pyramid, evaluate the HR manager’s proposal in Case 2. What is the
primary risk of their implementation strategy?...........................................................................36

4
Dark Patterns in UX Design
Dark patterns are deceptive design practices used in user interfaces to manipulate users into
taking actions they might not otherwise intend, such as hiding unsubscribe links.
Core Pattern Typology
 Dark Patterns: Deceptive practices that prioritize shareholder value over user well-
being.
 Anti-Patterns: Common designs that unintentionally lead to negative outcomes, such
as excessively lengthy subscription forms.
 Honest Patterns: Designs that prioritize transparency, such as showing upfront
pricing including taxes and shipping.
 Persuasive Patterns: Ethical techniques used to guide users toward desired actions,
such as prominent "Add to Cart" buttons or displaying genuine product reviews.
The Five Dark Strategies
Companies often utilize specific strategies to manipulate user behavior:
1. Nagging: Repeatedly asking or prompting a user to do something.
2. Obstruction: Making a process (like canceling a service) more difficult than it needs
to be.
3. Sneaking: Hiding or disguising relevant information.
4. Interface Interference: Manipulating the UI to prioritize certain actions over others.
5. Forced Action: Requiring a user to perform a specific action to access a feature.
Case Studies & Examples
 Epic Games (Fortnite): FTC settlement regarding unintended in-game purchases.
 TikTok: Fined €5 million by French regulators for its handling of user cookies.
 Amazon Prime: Lawsuit regarding the difficulty of canceling subscriptions compared
to the ease of signing up.
 Noom: $62 million settlement over automatic subscription renewals.

Data Privacy Laws & Regulations

Regulations provide the legal framework for data protection and define the rights of
individuals regarding their personal information.
Pakistan's Legal Landscape
 Pakistan Electronic Crimes Act (PECA) 2016: The primary legislation for
electronic crimes and data privacy.
 Personal Data Protection Bill 2025: An upcoming bill aimed at further codifying
data protection rights.
 Pakistan Telecommunications Re-organisation Act 1996: Requires operators to
ensure the privacy of customer data and prohibits unauthorized interception.
 Unsolicited Fraudulent and Obnoxious Communication Regulations 2009:
Provides protection from spam and establishes the "Do Not Call Register" (DNCR).
International Comparison: GDPR vs. CCPA
 Scope: GDPR applies globally to any business processing EU resident data, while
CCPA focuses on businesses processing California resident data.
 Consent: GDPR requires explicit, freely given consent; CCPA allows data collection
but requires an opt-out mechanism (e.g., "Do Not Sell My Personal Information"
link).
 Penalties: GDPR fines can reach €20 million or 4% of global revenue; CCPA fines
range from $2,500 to $7,500 per violation

5
Business ethics canvas
These notes provide a comprehensive summary of the research paper "Exploring the ethical
implications of business analytics with a business ethics canvas," focusing on the integration
of ethics into the analytics development process1.

1. Introduction & Context

The ethical aspects of data science and artificial intelligence (AI) have become major issues
as algorithms increasingly impact daily life, from medical treatments to financial
services2222. While algorithmic decision-making offers benefits—such as 86% accuracy in
detecting early-stage colorectal cancer—it also has a "dark side," including predictive
policing concerns and the Cambridge Analytica scandal333333333.
 Problem: Existing ethical work is often pitched at the level of abstract guiding
principles and fails to provide practical, grounded tools for practitioners4444.
 Regulatory Driver: The General Data Protection Regulation (GDPR) mandates
organizations address discrimination, bias, and provide a "right to an explanation"5.
 Proposed Solution: A Business Ethics Canvas (BEC) designed to be embedded in
the everyday exploration of analytics opportunities6.

2. Theoretical Foundations
The BEC is built upon the Markkula Center’s framework, which identifies five sources of
ethical standards7777:
Table 1: Markkula Ethical Framework & Business Analytics Implications
Ethics Ethical Questioning (Markkula, Implications for Business Analytics
Source 2018a) (adapted from Schwarz, 2005)
Does this action produce the most What are the benefits and harms of the
Utilitarian good and least harm for all analytics use case? Who benefits/is
affected? 8 harmed? 9
Does the action respect the dignity Whose rights are respected or infringed
Rights
and rights of all stakeholders? 10 by this analytics use case? 11
Does this action treat people equally How fair are the outcomes? Do they treat
Justice
or proportionally? 12 everyone the same or show favoritism? 13
What constitutes the common good in
Common Does this action best serve the
the community where the use case is
Good community as a whole? 14
developed? 15
Does this option lead me to act as the How does this use case define us as a
Virtue
sort of person I want to be? 16 person, company, or society? 17

Case Study: EuroTravel

The researchers used action research at EuroTravel, a European eCommerce travel
business18.

6
3. Business Model Canvas (BMC)
EuroTravel's business model (Fig. 1) relies on a computer-based booking platform that
provides unified booking across travel operators19.
 Value Propositions: Unified booking, travel budgeting, and travel experiences (e.g.,
cultural days out)20.
 Target Segments: Leisure, Business, and Commuter travelers21.
Selected Use Case: "DaysOut"
The project focused on "DaysOut," an analytics-driven recommendation engine: "As a user, I
want to know of attractive, feasible day trips for me so that I can have an adventure if I so
choose"22.

4. The Opportunity Canvas (OC)

7
Before ethical exploration, the team used an Opportunity Canvas (Fig. 2) to articulate the
business context23.
 Problems: Lack of awareness of adventure opportunities24.
 Solution Ideas: Personalized adventure trip recommendations25.
 Business Benefits: Increased revenue from ticket sales and ancillary services (hotels);
transformation of the customer relationship from transactional to experiential 26.

5. The Business Ethics Canvas (BEC)

The BEC (Fig. 3) depicts ethical concerns "surrounding" the solution and stakeholders to
allow for a clockwise workflow of exploration27.
Applying the BEC to "DaysOut" (Fig. 4)

8
The team used sticky notes to map opportunities (Green) and risks (Red)28282828:
 Utility (Opportunity): Creating more joy and adventure in people's lives29.
 Utility (Risk): Leaving customers with less money for essentials30.
 Common Good (Opportunity): Replacing car day trips with rail to reduce carbon
footprint31.
 Justice (Opportunity): Spreading wealth to less well-off communities32.
 Rights (Risk): Potential infringement on the community's rights to regulate traffic to
public assets33.

6. Process for Development

The BEC is implemented through a workshop-based approach:
1. Preparation: Identify stakeholders and print large-scale canvases34.
2. Solution Identification: Scribe the solution idea and all affected stakeholders using
yellow sticky notes35.
3. Ethical Exploration: Define positive opportunities (Green) and negative risks (Red)
for each ethical dimension using action-oriented language36363636.
4. Priority Voting: Participants use a limited number of votes to identify the most
consequential concerns37.
5. Iteration: Revise the Opportunity Canvas based on BEC findings (e.g., identifying
new business partnerships with local attraction owners)38.

9
Strategic Reflection

The research proposes an enhanced Business Analytics Methodology (BAM) (Fig. 5)39.
 Creative Tension: The Opportunity Canvas and Business Ethics Canvas act as
counterparts40. The OC provides the business rationale (value), while the BEC
provides stakeholders and ethical scrutiny (risks/governance)41.
 Beyond "Tick-Box" Ethics: By integrating the BEC into the development process,
ethics becomes a tangible exploration of value rather than a standard "rubber-
stamping" activity42.
 GDPR Alignment: The BEC directly addresses GDPR requirements through the
"Rights" (privacy, explanation) and "Justice" (removal of bias) boxes43.

10
Data governance & The Zachman Framework
This summary details the core concepts of data governance as presented in the provided slide
sets, focusing on the DAMA DMBOK framework, Aiken’s Pyramid, and the Zachman
Framework with its specific application to data ethics.

1. Data Governance Frameworks: DAMA DMBOK

The Data Management Body of Knowledge (DM-BOK), published by the Data

Management Association, identifies Data Governance as the central hub of an organization’s
data ecosystem1. It oversees 10 core functional areas:
 Data Architecture: Defines the structure of logical and physical data assets and
management processes throughout the data lifecycle2.
 Data Modeling and Design: The process of discovering, analyzing, and
communicating data requirements via data models3.
 Data Storage and Operation: Designing, implementing, and supporting data
storage4.
 Data Security: Ensuring proper data access while maintaining privacy and
confidentiality5.
 Data Integration and Interoperability: Designing and implementing data
movement and consolidation between sources6.
 Document and Content Management: Managing data stored in unstructured media7.
 Reference and Master Data: Maintaining core, critical shared data across the
organization8.
 Data Warehousing and Business Intelligence: Planning and controlling decision-
supporting data9.
 Metadata: Managing the information about the data in the organization10.
 Data Quality: Ensuring that data is "fit for use"11.

11
2. Peter Aiken’s Pyramid & Four Phases of Data Management

The DMBOK Pyramid illustrates that while Data Analytics is the "golden function" (most
value-added), it is only a small part of the ecosystem and requires a strong foundation of
other functions to work12.
Organizations typically progress through four phases:
 Phase 1 (Foundation): Begins with purchasing applications. Focus is on data
modeling, storage, and security13.
 Phase 2 (Data Quality): The organization addresses pains caused by bad
data. Reliable metadata and consistent data architecture become essential 14.
 Phase 3 (Strategic Integration): Data governance emerges to support structural
management and enables initiatives like data warehousing and master data
management15.
 Phase 4 (Advanced Analytics): Well-managed data finally enables high-value data
analytics and big data16161616.

12
13
Data Governance
Definition: The exercise of authority and control (planning, monitoring, and enforcement)
over the management of data assets.
Analogies & Metaphors
• Civil Government: Data governance is like a civil government. Laws do not determine
ethics, but they codify agreed-upon norms and provide a path for mediation/escalation when
disputes arise.
• The Kitchen Metaphor:
◦ Managing information is different from managing technology.
◦ The Scenario: A kitchen in a fancy restaurant.
◦ The Problem: You cannot deliver a good meal just by looking at the sharpness of knives
(Technology).
◦ The Reality: If ingredients are wrong or the menu isn't what was ordered, diners won't
care about the knives. Governance ensures the "ingredients" (data) and "menu" (strategy)
meet customer needs.
Governance Structure
Successful governance separates duties into two main views:
1. "Do the right things" (Legislative/Judicial): Focus on strategy and policy.
◦ Roles: Data Governance Council (DGC), Steering Committees.
2. "Do things right" (Executive): Focus on execution and oversight.
◦ Roles: Data Management Services (DMS), Data Stewards.
Key Tasks
• Shifting focus from IT to Information.
• Defining principles aligned with business strategy and ethos.
• Acting as an "honest broker" for internal data disputes.
--------------------------------------------------------------------------------
Data Stewardship
Definition: Accountability and responsibility for the use of data assets. A steward manages
data on behalf of others in the best interests of the organization.
The "Renting an Apartment" Analogy
• If you rent an apartment, you are a steward of that property.
• You do not own it, but your lease requires you to maintain it and not damage it.
• Similarly, data stewards hold data on trust for the benefit of someone else.
The 3DC Stewardship Framework
This framework categorizes stewards based on their relationship with data, rather than their
job title:
1. Doers: Work with data daily (operational level).
2. Definers: Define the meaning and purpose of information (strategic/tactical level).
3. Deciders: Make decisions on policies and rules (management/board level).
4. Co-ordinators: Facilitate discussion and act as honest brokers to resolve disputes.
--------------------------------------------------------------------------------
Data Architecture
Definition: Creating blueprints and specifications to define strategic data requirements and
align systems with business needs.
• The Librarian Metaphor: Data architects are like librarians who create and maintain the
organization’s knowledge about data structures and vocabulary.
• Four Architecture Perspectives:
1. Business: Value creation and capabilities.
2. Data: Definitions and relationships (e.g., what is a 'customer'?).

14
 3. Applications: Software structure and data consumption.
4. Technology: Physical environment/hardware.

The Zachman Framework

The Zachman Framework is a proactive tool used to classify an organization's architecture
and manage change17171717. It consists of a matrix organized by Interrogative
Columns and Perspective Rows.

Interrogative Columns (The Questions)

 What (Data): Business data, information, or objects18.
 How (Function): Business processes and operations19.
 Where (Network): Locations of business operations20.
 Who (People): Personnel, business units, and organizational hierarchy21.
 When (Time): Schedules, workflows, and process timings22.
 Why (Motivation): Goals, strategies, and the motivation for certain activities 23.
Perspective Rows (The Views)
 Planner (Scope Contexts): Defines the business purpose and strategy24.
 Owner (Business Concepts): Describes the organization where the system
functions25.
 Designer (System Logic): Outlines how the system satisfies information needs, free
from specific production constraints26.
 Implementer (Technology Physics): Shows how the system will be implemented
with specific technologies27.
 Sub-Constructor (Component Assemblies): Details implementation-specific system
elements28.
 User (Operations Classes): The functioning system in its actual environment29.

15
Core Rules: Each cell in the framework is unique and must align with the cells above, below,
and adjacent to it30. A single row provides a complete description of the enterprise from that
specific view31.

Applying Zachman to Data Ethics

The Zachman Framework can be used to explicitly embed ethics into data governance across
various perspectives32:
1. Definitions (What): Clearly define data artifacts and standards to ensure common
understanding33.
2. Procedures (How): Specify ethical considerations in the processes for collecting and
analyzing data34.
3. Responsibility (Who): Articulate roles responsible for ensuring ethical practices and
compliance35. Extend this to data access permissions—ensuring sensitive data is only
accessed by those who need it36.
4. Storage & Sharing (Where): Address ethical implications of secure storage,
transmission, and sharing37.
5. Temporal Ethics (When): Manage data timeliness, obsolescence, and disposal
policies38. Implement continuous improvement through regular reviews39.
6. Alignment (Why): Ensure analytics initiatives align with ethical values to prevent
misuse for financial gain40.

The Zachman Framework(summary Nllm)

An ontological model (6x6 matrix) describing the enterprise.
• The Construction Metaphor: Explains how different stakeholders view the same object
(e.g., a house) differently:
◦ Owner: Vision of the "finished product" and usage.
◦ Architect: Detailed design, trade-offs, and proportions.
◦ Builder/Trades: Plans for specific connections (taps, pipes, wires).
• Application: The framework links six communication interrogatives (What, How, Where,
Who, When, Why) to the steps needed to turn an abstract concept into reality.
Data Modelling
Definition: The process of discovering, analyzing, and scoping data requirements, then
representing them in a precise form (Data Model).
Model Levels
1. Conceptual: High-level business concepts (Executive view).
2. Logical: Attributes and relationships (Architect/Engineer view).
3. Physical: Database instantiation (Technician view).
Ethical Case Studies in Data Modelling
Data modelling is not just technical; it encodes ethics and bias.
• Case Study 1: Gender Representation
◦ Issue: Traditional "Male/Female" binary flags are insufficient for modern legal and
ethical standards.
◦ Solution: UK standards suggest three questions: Legal sex, Self-declared gender, and
Transgender status.
◦ Ethical Impact: Inclusivity and confidentiality must be "baked into" the model design.
• Case Study 2: Driver Safety App (Privacy Minimization)
◦ Scenario: An app tracks driving habits for rewards.
◦ Ethical Dilemma: Should the app upload GPS coordinates every 10 minutes to the
server? This creates a surveillance tool.

16
 ◦ Ethical Design: Calculate speed and distance locally on the phone and only upload the
result (points). This preserves privacy while achieving the business goal.
• Case Study 3: Voter Canvassing (Secrecy of the Ballot)
◦ Scenario: A political party wanted to integrate various data sources to analyze voting
patterns.
◦ Discovery: Modelling showed that combining specific discrete files would allow the
party to identify how individuals voted in past elections.
◦ Outcome: The client took a principled stand and stopped the data integration to protect
the constitutional right to a secret ballot.
--------------------------------------------------------------------------------
Data Quality Management
Definition: Activities ensuring data is "fit for consumption and business purposes".
• Ethical Principles (W. Edwards Deming):
1. Focus on the Customer: Understand their needs.
2. Drive out Fear: Blaming individuals for system errors is counterproductive.
3. Continuous Improvement: Don't rest on laurels.
• The "Hero Ethic" Trap: Relying on staff to manually fix data defects is unethical and
inefficient. The ethical approach is to fix the process that creates the defects.
• Root Causes: Often stem from management failures (e.g., poor software implementation,
lack of training) rather than front-line staff errors.
--------------------------------------------------------------------------------
Data Warehousing & Business Intelligence (BI)
• Goal: Integrating "data islands" into a single repository to support decision-making.
• Lifecycle Phases & Ethical Hazards:
◦ Acquisition: Are we obtaining data legally and transparently?
◦ Analysis: Risk: "If you torture data long enough, it will give you the answers you want."
Analysts must not bias results to fit a narrative.
◦ Action: Example: An automated system firing the lowest-performing staff based on
metrics without human review raises legal and ethical liability issues.
--------------------------------------------------------------------------------
Data Science and 'Big Data'
Difference from BI: BI is usually retrospective (what happened?), while Data Science is
predictive (what will happen?) or prescriptive (what should we do?).
• Big Data Attributes: Volume, Variety, Velocity.
• Key Technologies:
◦ Machine Learning: Algorithms that learn from data (Unsupervised, Supervised,
Reinforcement).
◦ Sentiment Analysis: Inferring emotion/attitude from text.
◦ Predictive Analytics: Forecasting future outcomes (e.g., share trading models).
Ethical Issues in Data Science
• The "Scientific Method" Gap: True science requires hypothesis testing and observation.
In business, Data Science is often conflated with BI, and the rigorous ethical codes of
experimenting on human subjects (common in academia) are often missing in corporate data
labs.
• Algorithmic Bias: Training data sets may contain historical biases that the algorithm learns
and perpetuates.
• The "Just Because We Can" Problem: The ability to infer private details (e.g., driving
habits, personal associations) does not mean an organization should do so.

17
Dama vs Zachman
Feature DAMA DMBOK Zachman Framework
Ontology & Classification: A schema
Process & Function: Defines the
for organizing and categorizing all
Primary Focus specific disciplines and activities
architectural artifacts of an
required for data management2222. 3333
enterprise .
The Wheel: 11 interconnected The Matrix: A $6 \times 6$ grid of
Structure knowledge areas with Data interrogative columns and perspective
Governance at the center4444. rows5555.
10 Functional Areas: Includes 36 Cells: Intersections of Interrogatives
Core
Data Security, Quality, Metadata, (What, How, etc.) and Perspectives
Components
and Architecture 6. (Planner, Owner, etc.)7777.
"How do we describe the
"How do we manage the data
enterprise?" Uses 6 interrogatives:
Key Questions asset?"Focuses on technical and
What, How, Where, Who, When,
operational execution8888.
Why9999.
Embedded in Practice: Ethics is a Artifact-Driven: Ethics is defined
neutral tool applied to disciplines within specific cells (e.g., "Who" for
Role of Ethics
like modeling or roles, "Why" for
quality101010101010101010. motivation)111111111111111111.
The Foundation: Viewed via
Alignment: Ensures analytics
Relationship to Aiken’s Pyramid;
initiatives align with business goals
Analytics governance/quality are
(Why) and roles (Who) 13131313.
prerequisites for analytics 12.
Central Hub: Acts as the authority Distributed: Governance is a specific
Governance
and control center for all data view (e.g., Owner/Executive) or
Role
activities14141414. interrogative (Who/Role)15.

18
Ethical issues in AI (case study)

19
Based on the provided document, here are in-depth, structured notes designed for learning
and extracting key concepts regarding the detection of future signals and the specific case
study on AI ethics.

1. Theoretical Background: Understanding "Future Signals"

This section defines the core concepts used to predict future trends.
1.1 Horizon Scanning and Weak Signals
 Horizon Scanning: A foresight methodology involving the analysis of external
environments to identify events, trends, and relationships.
 Weak Signals: These are the early warnings or "future-oriented information behind
future trends".
 Characteristics of Weak Signals:
o They are often new, surprising, and difficult to track amid noise.
o They typically have a substantial lag time before becoming mainstream.
o They represent opportunities to learn, grow, and evolve, or conversely, act as
threats.
1.2 Hiltunen’s Three Dimensions of a Future Sign
To measure a signal, Hiltunen (2008) proposed a 3D framework:
1. Signal: The visibility or number of signals (How often do we see it?).
2. Issue: The variety of events or the number of events (What is happening?).
3. Interpretation: The receiver's understanding of the meaning (Do we know what this
means?).
 Progression: A weak signal is low in all three dimensions (it is just starting to
diffuse). A signal becomes strongwhen the interpretation of its meaning becomes
clear and visibility increases.

2. Methodology: The Revised Analysis Framework

The authors argue that traditional methods rely too heavily on expert intuition (which is
costly and biased) or purely on quantitative text mining (which lacks context),. They propose
a Revised Framework.
2.1 The Quantitative Foundation (Yoon’s Approach)
The study builds on Yoon’s (2012) text mining method, which uses two key formulas to
measure signals:
 Degree of Visibility (DoV): Measures the frequency of a keyword (Represents the
"Signal" dimension).
 Degree of Diffusion (DoD): Measures the document frequency (Represents the
"Issue" dimension).
The Portfolio Maps:
 KEM (Keyword Emergence Map): Maps Term Frequency (TF) against DoV.
 KIM (Keyword Issue Map): Maps Document Frequency (DF) against DoD.
 Interpretation:
o Weak Signals: Low frequency/share but high growth rate (Quadrant A).
o Strong Signals: High frequency/share and high growth rate (Quadrant B).
2.2 The Qualitative Addition (The Authors' Contribution)
To fix the limitation where single keywords lack context or clear interpretation, the authors
added a qualitative step.
Process:
1. Collect Data: Web articles from EBSCO host (708 documents from 1990s–
2016),.

20
 2. Keyword Extraction: Use NLP to extract nouns/verbs/adjectives.
3. Map Construction: Create KEM and KIM based on statistics.
4. Topic Identification (New Step): Analyze co-occurrence (words appearing
together) to identify specific "Issue Topics" rather than just isolated words,.
5. Classification: Categorize topics into Strong, Weak, Well-known, or Latent
signals.

3. Case Study: Ethical Issues in Artificial Intelligence

The authors applied their framework to analyze how society discusses ethical problems
regarding AI.
3.1 Context and Data Trends
 The Rise of AI: Technologies like Google's AlphaGo and self-driving cars are rapidly
commercializing,.
 The Problems: Incidents like the Palo Alto security robot injuring a child,
Microsoft’s "Tay" chatbot learning racism, and Tesla autopilot accidents have raised
ethical concerns,.
 Document Trend: There was a dramatic increase in documents regarding AI ethics
starting in 2014.
3.2 Classification of Ethical Signals (Key Findings)
The study identified 11 specific topics and classified them into four categories based on their
signal strength.
A. Strong Signals
Definition: High attention, high growth rate. The problems are understood, and stakeholders
are taking action (e.g., laws/regulations).
Topic Keywords Definition
Camera, security, Infringement and leakage of personal info by
Privacy
surveillance, drone AI/automated surveillance.
Car, self-driving, accident, Controversy over accident risks and ethical
Safety
vehicle algorithms in autonomous vehicles.
Human Animal, abuse, crime, Malicious learning by AI, use of AI in crime, or
Abuse violence violence against robots.
B. Weak Signals
Definition: Potential future problems. Applications are not fully realized in daily life yet;
impacts are discussed mostly by experts, not the general public.
Topic Keywords Definition
Legal Legal, responsibility, Disputes over who is responsible (locus/scope) for
Liability conflict accidents caused by AI.
War, military, attack, Development of combat robots with autonomous
Killer Robots
drone decision-making capabilities.
Humanity, destroy, Disputes over the dystopia where AI transcends
Singularity
Hawking, fear human ability.
C. Well-Known but Not Strong Signals
Definition: High share of discussion but low growth rate. These are vague fears or long-
standing fictional tropes (Sci-Fi) without specific direction.
Topic Keywords Definition
Imaginary
Race, Sci-Fi, scenario Vague fears based on movies and comics.
Fear
Creature, synthetic, Confusion of identity caused by the appearance of
Identity
humanoid humanoid robots.
21
D. Latent Signals
Definition: Low share and low growth rate. Potential signals that may vanish or simply
persist without impact,.
Topic Keywords Definition
Final "Big red button",
Systems to stop runaway AI technology.
Safeguard defense system
Emotional Emotion, mimic, heart,
Side effects of robots that simulate human emotion.
Robot child
Decision by Surgery, doctor, expert, Socio-economic issues of robots supporting/making
Robot replicate high-level decisions.

4. Conclusion and Implications

 Methodological Value: The revised framework allows for an "ambidextrous
approach," balancing the rigor of quantitative big data analysis with the flexibility of
qualitative interpretation.
 Policy Value: This method helps governments and businesses prioritize which issues
require immediate intervention (Strong Signals like Privacy/Safety) versus which
require monitoring for future consensus (Weak Signals like Legal Liability).
 Limitations: The study uses keyword co-occurrence for interpretation; future
research should use semantic network analysis to better understand the structural
relationships between words over time,.

Analogy for Understanding

Think of this methodology like Weather Forecasting:
 Traditional Method: An old sailor looking at the sky (Expert Intuition). It is useful
but limited to their personal view.
 Yoon’s Quantitative Method: A barometer measuring pressure changes
(Data/DoV/DoD). It gives raw numbers but doesn't tell you why the pressure is
dropping.
 The Revised Framework (This Study): A meteorologist looking at the
barometer and satellite imagery to identify the specific storm formation (Quantitative
+ Topic Identification). They can tell you not just "pressure is low," but "there is a
Hurricane named 'Privacy' approaching."

Building Transparency into AI Projects

Core Thesis
Transparency in AI is not just a buzzword but a systematic chain of communication across all
stages of development and deployment. It is essential for building trust, mitigating risk, and
ensuring ethical AI use. It goes beyond just explaining how a model works ("explainability")
to encompass the entire process—from design to deployment and monitoring.

1. The Problem & A Cautionary Tale

Example: A major tech company faced backlash in 2018 for an AI that impersonated a
human to make restaurant reservations, including using "ums" and "ahs." The public felt
deceived.
Key Insight: Transparency is demanded by consumers and stakeholders. It involves
communicating:
When an AI is being used.
Why it was chosen.
22
 How it was designed, developed, and deployed.
How it is monitored and retired.

---

2. The Four Impacts of Transparency

Building a transparent process delivers concrete benefits:

1. Decreases Risk of Error & Misuse: Poor communication between teams leads to failure.
Example: An AI for detecting tumors in X-rays was designed to be highly sensitive (low
false negatives), leading to many false positives. Radiologists weren't told this, so they spent
excessive time reviewing AI-flagged scans, thinking they'd missed something. Transparency
would have allowed them to use the tool efficiently.

2. Distributes Responsibility: Informed stakeholders can make responsible decisions and be

held accountable.
Executives need to know design choices to approve deployment.
Users need to know how to interpret outputs correctly.
Without transparency, blame falls on those who withheld information or failed to create a
culture of open communication.

3. Enables Internal & External Oversight: Critical for catching ethical, legal, and reputational
risks.
Data scientists shouldn't be the sole decision-makers for high-risk systems.
Regulators and internal risk managers (e.g., boards) cannot do their jobs if the process is
opaque.

4. Expresses Respect for People: Preserves human autonomy and prevents manipulation.
Using AI without disclosure (like the reservation AI) is deceptive.
Transparency allows people to make informed choices, respecting their right to self-
determination.

3. How to Implement Transparency: A Practical Guide

Transparency is not "all or nothing" but must be strategically implemented.

1. Identify Stakeholders: Map all internal and external parties (e.g., data scientists,
executives, end-users, regulators, customers).
2. Determine Their Needs: Ask what each stakeholder needs to know to do their job
effectively and make informed decisions.
3. Balance Transparency with Risk: Weigh the need to share information against risks like
exposing intellectual property. High-risk AI applications demand greater transparency.
4. Tailor the Communication:
Explanations must be adapted to the audience's technical knowledge (CIO vs. CEO vs.
consumer).
Choose the right medium (email, training, reports) for clear and effective knowledge
transfer.
5. Systematize into Policy: Formalize decisions about what to communicate, to whom, and
when into company policies and procedures.

23
Key Distinction: Transparency vs. Explainability
Explainability: Focuses on the internal mechanics of a model—how inputs are transformed
into outputs (the "how").
Transparency: Encompasses the entire process around the model—the decisions, data,
design, and deployment context (the "what," "why," and "who").
Both are crucial for building trust, and the principles of good communication apply to both.

Conclusion
AI transparency is a concrete, operational requirement—not a theoretical ideal. It requires
intentional design of communication chains across the AI lifecycle to reduce risk, enable
accountability, and respect the autonomy of everyone affected.

Conversations Companies Need to Have":

Core Problem
While concerns about AI ethics (bias, privacy, opacity) are now mainstream, translating these
concerns into concrete, actionable risk-mitigation strategies within companies remains a
significant challenge.

Proposed Solution: A Structured, Conversational Framework

Companies must move beyond abstract worry by forming a dedicated senior team and having
three specific, progressive conversations.

1. Form the Right Team

Assemble a senior-level working group with four essential roles:
Technologists: To assess what is technologically feasible.
Legal/Compliance Experts: To align plans with existing regulations and anticipate new
ones.
Ethicists: To systematically identify ethical and reputational risks beyond mere compliance.
Business Leaders: To ensure strategies align with business goals and economic feasibility.

2. The Three Crucial Conversations

Conversation 1: Define Your Ethical Standards

Go beyond legal compliance (e.g., GDPR) and establish organization-specific ethical
benchmarks. Answer tough questions:
What constitutes a "discriminatory" model? Is the benchmark "better than historical human
bias" or an absolute standard?
When are "black box" models acceptable? What level of explainability is required?

Outcome: Clear, actionable frameworks and due diligence processes for product teams.

Conversation 2: Identify the Gaps

Assess the distance between your current state and your new standards. Recognize that
technical tools (e.g., fairness metrics, explainability software like LIME/SHAP) are necessary
but not sufficient. The team must ask:
What specific risk are we mitigating?
How do quantitative tools help, and what gaps do they leave?

24
 What qualitative assessments are needed to fill those gaps, and who should do them?

Outcome: A clear roadmap for integrating quantitative and qualitative approaches.

Conversation 3: Understand the Sources & Operationalize

Move beyond simplistic explanations like "biased data." Deeply analyze the complex sources
of ethical risks:
How exactly did the training data become biased?
How were model inputs weighted, thresholds set, or objective functions chosen?
Understanding these root causes is essential for selecting effective, tailored mitigation
strategies.

Outcome: Effective, operational solutions that address the true complexity of AI ethics
problems.

Key Takeaway
AI ethical risk is concrete, not theoretical. Effective management requires a multidisciplinary
team and a structured process to define standards, diagnose gaps, and implement deeply
informed solutions.
Here are separate summaries for each of the provided documents:

Summary: "Google Let OpenAI Scrape YouTube Data Because

Google Was Doing It Too" (PCMag)
This article reports on allegations that major tech companies (OpenAI, Google, and Meta)
have engaged in or considered ethically and legally questionable data collection practices to
train their AI models, driven by a scarcity of high-quality training data.

Key Points:
The Allegation: The New York Times reported that OpenAI transcribed over a million hours
of YouTube videos to train its GPT-4 model, potentially violating YouTube's terms of service.
Google's Complicity: Google, which owns YouTube, allegedly engaged in similar data
scraping from its own platform. This is suggested as a reason Google did not aggressively
pursue action against OpenAI.
Contradictory Statements: YouTube's CEO stated that scraping transcripts or video data
violates its terms of service but was evasive when asked directly if OpenAI had done so.
Broader Context:
Data Hunger: The article underscores a "data arms race," with AI leaders like Sam
Altman warning that high-quality data will eventually run out.
Other Examples: Meta reportedly considered acquiring publisher Simon & Schuster or
scraping data preemptively, hoping to avoid lawsuits. Google also changed its terms of
service to allow using public user data from Docs and Maps for AI training.
Legal Backlash: The New York Times has sued OpenAI and Microsoft for using its
copyrighted content, highlighting the growing legal battles over AI training data.

25
Core Issue:
The pursuit of massive datasets for AI advancement is pushing companies into legal gray
areas, raising significant questions about copyright, terms of service, and corporate
accountability.

Summary: "Machine Bias" (ProPublica)

This investigation exposed severe racial bias in algorithmic risk assessment tools widely used
in the U.S. criminal justice system to predict the likelihood of a defendant reoffending.

Key Findings (from the original report):

The Tool: The article focused on COMPAS (Correctional Offender Management Profiling
for Alternative Sanctions), used in states like Florida to guide sentencing and parole
decisions.
ProPublica's Analysis: Their investigation revealed the algorithm was biased against Black
defendants.
Black defendants were far more likely than white defendants to be incorrectly labeled as
higher risk (more false positives).
White defendants were more likely than Black defendants to be incorrectly labeled as
lower risk (more false positives).
Impact: These flawed risk scores could lead to harsher sentences, denied parole, and
increased bail for Black defendants, perpetuating systemic inequities.
The "Black Box" Problem: The proprietary algorithm's exact logic was not transparent,
making it difficult to challenge or understand its biased outcomes.

Core Conclusion: The report demonstrated that algorithms can automate and amplify
existing societal biases, posing a grave threat to justice and fairness when deployed
without rigorous auditing for discrimination.
Here is a concise summary of the HBR article "Overcoming the C-Suite's Distrust of AI":

Overcoming the C-Suite's Distrust of AI

The Core Problem
While AI is widely adopted for tactical, low-level decisions (e.g., credit scoring, chatbots),
the C-suite remains deeply distrustful of using it for high-level strategic decisions (e.g.,
corporate strategy, product lines, major partnerships). Surveys show a majority of executives
are uncomfortable relying on AI/analytics, preferring their own intuition and experience.

Root Causes of Distrust

Executives' hesitancy stems from legitimate concerns about:
1. Unreliable Models: AI projects often fail due to poor data quality, outdated information, or
models that can't adapt to new realities (like a pandemic).
2. Data Bias: AI trained on biased data will produce discriminatory recommendations,
leading to ethical and legal risks.
3. Unethical Outcomes: Fear that AI-assisted decisions may not align with corporate moral
values, exposing the company to liability.
4. Black Box Decisions: Lack of transparency and explainability makes it impossible to
understand or justify high-stakes AI recommendations.

---
26
Four Actions to Build Executive Confidence

1. Create Reliable AI Models:

Invest in data quality. Data scientists spend ~50% of their time on data preparation
(cleansing, labeling).
Use tools for synthetic data, debiasing, and cleansing to provide timely, accurate, and
context-rich data.

2. Avoid Data Biases:

Proactively vet and cleanse data of racial, gender, social, and ideological biases
("garbage in, garbage out").
Assign lower confidence scores to data from unvetted sources and focus preprocessing
on controlling discrimination.

3. Ensure Ethical & Moral Decisions:

Integrate ethical frameworks (like Stuart Russell's Value Alignment Principle) that
"reward" AI for acceptable behavior.
Implement checks and balances to ensure AI-assisted decisions reflect corporate values.

4. Explain AI Decisions:
Move beyond "black box" outputs. AI decisions involving major resources or people's
lives must be explainable.
Maintain accountable data practices (secure, auditable storage) and use third-party
governance.
Keep humans in the loop—AI provides probabilistic recommendations, not absolute
certainties, requiring human critical thinking and oversight.

Key Takeaways & Tasks for Leaders

Promote ownership of AI ethics beyond the IT department.
Recognize AI as a decision-support tool, not an oracle; critical thinking remains essential.
Target AI to high-impact areas first to demonstrate value.
Ensure fairness through transparency and regular audits of AI results.
Take responsibility for decisions; don't blame the AI.

Conclusion
AI's role in the C-suite is augmented intelligence—assisting human leaders with reliable,
unbiased, explainable, and ethical insights. Building trust requires demonstrating that AI
processes are robust and transparent, allowing executives to confidently incorporate them into
strategic decision-making.

Summary of ethics and AI files

Case study Core Focus

A case study on ethical issues in Foresight methodology using text mining to

1
[Link] detect emerging ethical signals in AI.

27
 Case study Core Focus

Ethics and AI: 3 Conversations Corporate governance framework for

2
Companies Need to [Link] structuring AI ethics conversations.

3
Building transparency into AI Operational transparency as a chain of
&
(Parts 1 & 2).pdf communication across the AI lifecycle.
4

google let openai scrape youtube Real-world data ethics scandal involving AI
5
data [Link] training data sourcing.

Investigation into algorithmic bias in

6 Machine [Link]
criminal justice risk assessment tools.

Overcoming C suites distrust of Strategies to build executive trust in AI for

7
[Link] strategic decision-making.

Ethics and AI Slides

Part 1: Building Transparency into AI Projects

(Based on Files 3 & 4: "Building transparency into AI")
 Core Idea: Transparency is not a one-time disclosure but a continuous, systematic
chain of communication across the entire AI lifecycle.
Introduction:
o Starts with the example of a human-like robot making reservations—an act
of deception that sparked public backlash. This real-world case (referenced in
the original article) illustrates the growing demand for transparency.
What is Transparency in AI?
o It's more than just explaining how a model works. It requires disclosing:
 Why the AI was built.
 How it was designed and tested.
 On what grounds it was deployed.
 How it's monitored and maintained.
o It is a "systematic transference" of knowledge between all stakeholders,
from data collectors to executives to end-users.
Impacts of Being Transparent (The Four Key Benefits):
1. Decreases Risk of Error and Misuse: Prevents scenarios like the X-ray AI
case (from the article) where radiologists misused a tool because they weren't
told it was calibrated for high sensitivity.
2. Distributes Responsibility: Enables informed decision-making. Executives
can be held accountable for approving a model only if they were transparently
informed of its design and limitations.
3. Enables Internal & External Oversight: Allows regulators (like in the bank
investigation example) and internal risk boards to perform their duties,
catching ethical or legal issues technologists might miss.

28
 4. Expresses Respect for People: Informing users they are interacting with an
AI respects their autonomy and prevents manipulation, contrasting with the
deceptive reservation robot.
Implementing Transparency (What Good Communication Looks Like):
o Balance is Key: Transparency is not "all or nothing." It must be weighed
against risks like exposing intellectual property.
o Identify Stakeholders: Map everyone involved (internal and external) and
determine what they need to know to do their job or make informed choices.
o Tailor Explanations: Communication must be adapted to the audience (e.g.,
technical details for a CIO, plain-language summaries for consumers).
Crucial Distinction: Transparency vs. Explainability:
o Explainable AI: Focuses on the internal mechanics—how inputs become
outputs.
o Transparency: Encompasses the entire external process—the what, why,
who, and whenaround the model.
o Both are essential for building trust.

Part 2: Overcoming the C-Suite's Distrust of AI

(Based on File 7: "Overcoming C suites distrust of AI")
 Core Problem: While AI is trusted for tactical decisions (e.g., chatbots), the C-suite
deeply distrusts it for strategic decisions (e.g., corporate strategy), preferring gut
instinct.
Four Actions to Build Executive Confidence:
1. Create Reliable Models: The 1 cause of AI failure is poor data quality.
Executives need models built on timely, clean, and relevant data. Data
scientists spend ~50% of their time on preparation.
2. Avoid Data Biases: "Garbage in, garbage out." If training data is biased (e.g.,
containing racial or gender bias as in the ProPublica "Machine Bias" case
[File 6]), the AI's recommendations will be too. This must be actively
cleansed.
3. Ensure Ethical/Moral Decisions: AI must align with corporate values.
Mentions importing research like Stuart Russell's Value Alignment
Principle, which trains AI on stories to learn acceptable behavior.
4. Explain AI Decisions (Move Beyond Black Box): For high-stakes decisions,
saying "the AI decided" is insufficient. Processes must be in place to audit
and explain outcomes. Humans must remain in the loop to override AI.
Key Realities Highlighted:
o Probabilistic vs. Deterministic: AI deals in probabilities, not certainties. A
98% confidence score in healthcare isn't enough.
o Human Oversight is Non-Negotiable: Executives must retain the ability to
question and override AI, especially in complex, unpredictable scenarios.
Tasks to Raise Executive Confidence:
o Promote ownership of AI ethics beyond the IT department.
o Use critical thinking to challenge AI output.
o Ensure fairness through transparency.
o Audit regularly and take responsibility for decisions—don't blame the AI.

29
Part 3: Ethics and AI: 3 Conversations Companies Need to Have
(Based on File 2: "Ethics and AI_ 3 Conversations Companies Need to Have")
Prerequisite: Form the Right Team.
o A senior-level, multidisciplinary working group is essential:
 Technologist: Assesses feasibility.
 Legal/Compliance Expert: Navigates regulations.
 Ethicist: Identifies ethical/reputational risks.
 Business Leader: Aligns with business goals.
The Three Conversations:
1. Define Your Ethical Standard: Go beyond legal compliance (e.g., GDPR).
Ask tough, company-specific questions: "What do we consider a
discriminatory model?" or "When is a 'black box' AI acceptable?"
2. Identify the Gaps: Audit the difference between your current state and your
new ethical standard. Recognize that technical tools (like fairness software)
are necessary but not sufficient. The team must determine what qualitative
assessments are needed to fill the gaps.
3. Understand Problem Sources & Operationalize Solutions: Move beyond
vague blame like "biased data." Deeply analyze the root causes (e.g., how
data became biased, how model weights were set) to design effective, tailored
mitigation strategies.

TOGAF
TOGAF, which stands for The Open Group Architecture Framework, is the world’s most
widely used framework for Enterprise Architecture (EA).1
Think of it as a detailed "blueprint" or a "recipe book" that helps large organizations align
their business goals with their IT infrastructure. It ensures that when a company builds new
software or changes its processes, everything works together smoothly rather than becoming
a disorganized mess of incompatible systems.

1. The Four Architecture Domains (BDAT)2

TOGAF breaks down an entire organization into four key "layers," often referred by the
acronym BDAT:3
 Business Architecture: Defines the business strategy, governance, and key business
processes.4 It answers: "What is the business trying to achieve?"
 Data Architecture: Describes the structure of an organization's logical and physical
data assets.5 It answers: "What information do we need and where is it kept?"6
 Application Architecture: Provides a blueprint for the individual applications to be
deployed and how they interact.7 It answers: "Which software systems do we need to
process our data?"8
 Technology Architecture: Describes the hardware, software, and network
infrastructure needed to support the other three layers.9 It answers: "What servers and
networks are running everything?"

2. The Architecture Development Method (ADM)10

The "engine" of TOGAF is the ADM.11 It is a circular, iterative process used to create and
manage an organization's architecture.12 Instead of a "one-and-done" project, it is a
continuous cycle of improvement.13
The 10 Phases of the ADM Cycle:

30
 1. Preliminary: Preparing the organization and defining "how" we will do
architecture.14
2. Phase A: Architecture Vision: Defining the scope and getting buy-in from
stakeholders.15
3. Phase B: Business Architecture: Mapping out business processes.16
4. Phase C: Information Systems Architecture: Designing the Data and
Applications.17
5. Phase D: Technology Architecture: Defining the underlying tech stack.18
6. Phase E: Opportunities & Solutions: Identifying the actual projects to be built.19
7. Phase F: Migration Planning: Creating a roadmap to move from the "Old Way" to
the "New Way."20
8. Phase G: Implementation Governance: Overseeing the actual build to ensure it
follows the blueprint.21
9. Phase H: Architecture Change Management: Monitoring for new tech or business
changes.22
10. Requirements Management: A central process that feeds into every other phase to
ensure needs are met.23

3. Key Concepts & Components

Beyond the ADM, TOGAF includes several "toolkits" to help architects:
 The Enterprise Continuum: A "virtual repository" of all architectural assets
(models, patterns, descriptions) ranging from generic industry standards to
organization-specific designs.24
 Architecture Repository: The actual place (database or folder) where an
organization stores its blueprints and standards.
 Building Blocks: Reusable components.25 For example, a "Login Service" could be
an Architecture Building Block (ABB) that is reused across multiple applications.

4. Why Organizations Use TOGAF

 Common Language: It provides a standardized vocabulary so that IT, Management,
and Stakeholders all understand each other.26
 Efficiency: By using reusable building blocks, companies save time and money on
new projects.27
 Risk Reduction: A structured approach ensures that security and compliance are
built-in from the start, not added as an afterthought.28
 Flexibility: While it is a massive framework, it is designed to
be modular.29 Organizations can "pick and choose" only the parts of TOGAF that
they actually need.30

5. What’s New in the 10th Edition?

The latest version (released as The TOGAF Standard, 10th Edition) is designed for the
modern era.31 It is more Agile-friendly and provides better support for:
 Digital Transformation: Helping companies move to the cloud.32
 Microservices & DevOps: Integrating architecture with modern software delivery.33
 Modularity: The core standard is now smaller, with "Series Guides" that provide
specific advice for different industries or technologies.34

31
TOGAF VS ZACHMAN VS DAMA-DMBOK
To help you compare these three frameworks, it is important to understand that they are not
competitors. Instead, they operate at different levels of the organization: TOGAF is the
process, Zachman is the blueprint, and DAMA-DMBOK is the specialized technical manual
for data.
Comparison Table: TOGAF vs. Zachman vs. DAMA-DMBOK
Dimension TOGAF (10th Ed.) Zachman Framework DAMA-DMBOK (v2)
Methodology (A "How- Ontology / Taxonomy (A Body of Knowledge (A
Core Nature
To" process). "What-Is" classification). "Best Practices" guide).
Categorizing and
The cycle of designing, Strategic management and
Primary organizing every
planning, and governance of data as an
Focus perspective of an
implementing EA. asset.
enterprise.
The ADM Cycle (A The 6x6 Matrix (36 cells The DAMA Wheel (11
Visual
circular flow of 10 of perspectives vs. data management
Symbol
phases). questions). knowledge areas).
Complete: Every Deep: Specifically
Broad: Business, Data,
possible view of the focused on the Data
Scope Application, and
organization from CEO to domain and Data
Technology (BDAT).
Coder. Governance.
"Have we documented
Key "How do we build the "How do we manage and
every part of the
Question architecture?" protect our data assets?"
architecture?"
Strategic roadmaps, A set of "primitive" Data governance policies,
Main Output architecture definitions, models that describe the data models, and quality
and migration plans. enterprise. standards.
Enterprise Architects, IT Business Analysts, Data Governance Officers,
Target
Leaders, Solution Architects, and Systems CDOs, Data Architects,
Audience
Architects. Designers. DBAs.
Process-driven, Holistic and neutral; Deep technical expertise
Strengths repeatable, and globally ensures no stakeholder and standards for the data
recognized standard. view is missed. lifecycle.
Can become overly Provides no instructions Limited guidance on
Weaknesses bureaucratic if not on howto actually build business or technology
tailored correctly. anything. architecture (Non-data).

How They Work Together (The Synergy)

Most mature organizations do not choose one; they use all three in a "stacked" approach:
1. Zachman as the Map: You use the Zachman matrix to ensure you have identified all
stakeholders (Rows) and answered all their basic questions (Columns). It acts as
your inventory to make sure there are no gaps in your knowledge.
2. TOGAF as the Engine: You use the TOGAF ADM cycle to actually move the
project forward. When you reach "Phase C" (Information Systems Architecture), you
realize you need to build the data layer.
3. DAMA-DMBOK as the Toolkit: To perform the work in Phase C effectively, you
pull out DAMA-DMBOK. It provides the specific "how-to" for data quality, metadata
management, and security that TOGAF only mentions at a high level.
Summary Analogy

32
  Zachman is the blueprint cabinet that ensures you have a drawing for the plumbing,
the wiring, the foundation, and the roof.
 TOGAF is the General Contractor’s manual that tells you the order in which to
pour the concrete, frame the walls, and install the roof.
 DAMA-DMBOK is the Master Plumber’s handbook that tells you exactly what
size pipes to use and how to ensure the water pressure is correct.

Case studies
Scenario:
During an upcoming meeting, employees are allowed to present and discuss innovative ideas.
In order to bolster creativity, successful ideas are given bonuses. One such idea is a business
analytics use case.
Business analytics use case 1:
An employee from the marketing department suggests using advanced sentiment analysis to
improve marketing campaigns. By analyzing customer interactions on social media, emails,
and support calls, the Al system would identify trends and sentiments, allowing the marketing
team to tailor their campaigns more effectively. The goal is to enhance customer engagement
and drive sales growth.
While the idea is met with enthusiasm, several ethical concerns arise. The use of Al to
analyze private customer communications raises significant privacy issues, as customers may
not be aware that their interactions are being scrutinized. Additionally, there are concerns
about the potential biases in the sentiment analysis, which could lead to inaccurate or unfair
targeting of customers. The cybersecurity team highlights the importance of protecting the
vast amounts of data being collected and ensuring compliance with data protection
regulations. The marketing team must balance the potential benefits öf enhanced customer
engagement with the need to address these ethical, privacy, and security challenges.
Relevant Topic
Explanation & Citation
from Document
Data Science and This case utilizes Sentiment Analysis (inferring emotion from text)
'Big Data' and Machine Learningto process high volumes of interaction data1111.
The scenario mentions analyzing private communications, which raises
Data Privacy & significant privacy issues2. This links to Pakistan's Personal Data
Legal Landscape Protection Bill 2025 and the requirement for explicit consent
under GDPR3333.
The conflict between "tailoring campaigns" (Utility/Opportunity) and
Business Ethics
"scrutinizing interactions without awareness" (Rights/Risk) is a perfect
Canvas (BEC)
application for the BEC4444.
The concern about "potential biases in sentiment analysis" leading to
Data Quality
unfair targeting relates to ensuring data is "fit for use" and avoiding
Management
discriminatory outcomes5555.
Zachman Applying Zachman to data ethics, this case must address the "Why"
Framework (Alignment) to ensure the initiative aligns with ethical values rather
(Motivation) than just financial gain6.
Business analytics use case 2:
Another from the HR department highlights the less impressive learning curves, low
satisfaction rates and higher than industry average turnover. He stresses that if there are
underlying issues that are being ignored, they could lead to unexpected negative outcomes.

33
Therefore, it is important to explore and be informed. He suggests using analytics and Al in a
way similar to that suggested by the marketing employee, but for 'internal marketing'
purposes. Employee interactions, emails, messages through teamworking apps, support calls,
HR complaints, HR requests as well as comments filled out in employee satisfaction surveys
and performance reviews would all be used. In order to protect privacy, this would be done
through Al. This would allow them to conduct an internal sentiment analysis of their
employees, and be informed of underlying issues etc. This information could also be used to
suggest HR strategies, or boost productivity. To keep costs low, they would opt for more
affordable Al models being offered, which do not come with built-in cybersecurity features.
While important confidential and proprietary data is protected by a good security system, the
HR manager is informed that it is priced in a way that if it is employed to protect all this data,
the cost would be prohibitive. It is then suggested to protect this by using a common open-
source method that is known to not be as strong as more recent proprietary methods but is
better than no protection.
Relevant Topic
Explanation & Citation
from Document
The document defines "Human Abuse" in AI ethics as malicious
Ethical Issues in learning or infringing on human dignity; monitoring private employee
AI (Human Abuse) communications through teamworking apps closely mirrors this
concern7.
The decision to use "common open-source methods" known to be
Data Security &
weaker due to cost constraints directly challenges the Data
Governance
Security functional area of DAMA-DMBOK8.
The HR manager is acting as a "Decider" or "Steward"9. Under
the "Renting an Apartment" analogy, they have a responsibility to
Data Stewardship
maintain and protect the data asset for the benefit of the
organization10101010.
Building The "internal marketing" purpose of the AI requires Transparency. The
Transparency into C-suite or employees need to know "Why" the AI was chosen
AI and "How" it is monitored to avoid a "Black Box" distrust111111111111111111.
This scenario specifically touches on Justice (treating people
Markkula Ethical
fairly/proportionally) and Rights (whose rights are infringed by this use
Framework
case)12.
Common Themes Across Both Cases
Both scenarios are deeply related to the "3 Conversations Companies Need to Have":
 Conversation 1 (Standards): Defining what constitutes a "discriminatory" or "overly
invasive" model.
 Conversation 2 (Gaps): Identifying the lack of robust cybersecurity in Case 2 as a
critical gap.
 Conversation 3 (Operationalize): Understanding how to mitigate bias in sentiment
tools before they are deployed

Question 1: The Ethics of Trade-offs

"In Case 2, the HR manager opts for a weaker open-source security method to save costs.
Using the DAMA-DMBOK framework and the 'Renting an Apartment' analogy, explain why
this is an ethical failure of data stewardship."
Answer:

34
  Data Stewardship Definition: Stewardship is the accountability and responsibility
for the use of data assets on behalf of others.
 The Analogy: Just as a tenant who rents an apartment is required by their lease to
maintain it and not damage it, a data steward holds data "on trust" for the benefit of
the organization.
 The Failure: By choosing a security method known to be "not as strong," the HR
manager (the "Decider") is failing to protect the asset they are entrusted with.
 DAMA Alignment: DAMA places Data Governance at the hub to ensure authority
and control; opting for weak security ignores the functional requirement of Data
Security, which is essential to maintain privacy and confidentiality.

Question 2: Applying the Business Ethics Canvas (BEC)

"For Case 1, identify one 'Green' Opportunity and one 'Red' Risk using the Business Ethics
Canvas categories of 'Utility' and 'Rights'."
Answer:
 Utility (Green Opportunity): Enhancing customer engagement and driving sales
growth by identifying trends and tailoring campaigns more effectively.
 Utility (Red Risk): If the sentiment analysis is biased, it could lead to inaccurate or
unfair targeting, potentially harming the company's reputation or the customer's
experience.
 Rights (Red Risk): The "scrutiny" of private communications (emails/calls) without
explicit awareness infringes on the customer's right to privacy and self-determination.

Question 3: Building Executive Trust

"The C-Suite is often distrustful of AI. Based on the HBR article 'Overcoming the C-Suite's
Distrust of AI,' identify two root causes of distrust present in Case 1 and suggest how to build
confidence."
Answer:
 Root Cause 1: Data Bias: The marketing team is worried about "potential biases in
the sentiment analysis". To build confidence, they must proactively vet and cleanse
data of ideological or social biases.
 Root Cause 2: Black Box Decisions: Customers and stakeholders may not
understand how interactions are being "scrutinized".
 Solution: Move beyond "black box" outputs by ensuring Explainability (how the
model works) and Transparency (why the data is being used). The company should
also "keep humans in the loop" to oversee high-stakes marketing decisions.

Question 4: Strategic Alignment (Zachman Framework)

"How can the Zachman Framework be used to address the 'Legal Liability' weak signal in
Case 2?"
Answer:
 Identifying the Weak Signal: Legal liability for accidents or data breaches caused by
AI is currently a "weak signal" because the impacts are discussed by experts but not
yet fully realized in daily life.
 Using Zachman (Who/Role): The framework requires articulating exactly which
roles (e.g., HR, IT, or the AI vendor) are responsible for ensuring ethical practices and
legal compliance.

35
  Using Zachman (How/Process): The organization must specify the ethical
considerations and legal safeguards within the actual processes used to collect
employee data.
 Using Zachman (Why/Motivation): This ensures the "internal marketing" initiative
aligns with the organization's actual ethical standards rather than just short-term cost-
cutting on security.
Question 5: Ethical Governance and "The Three Conversations"
Q: In Case 1, the cybersecurity team highlights the importance of protecting data. Which of
the "Three Crucial Conversations" must the senior team have before deploying this AI?
Answer:
 Conversation 1 (Define Ethical Standards): The team must move beyond legal
compliance and define what constitutes a "discriminatory" model and when "black
box" sentiment analysis is acceptable.
 Conversation 2 (Identify Gaps): They must assess the distance between their current
practice and their standards, recognizing that technical fairness metrics are not
sufficient on their own.
 Conversation 3 (Understand Sources): They must analyze the root causes of
potential bias—such as how the social media training data was weighted or how
thresholds for "negative sentiment" were set.
Question 6: The Foundations of Data Management
Q: Using Aiken’s Pyramid, evaluate the HR manager’s proposal in Case 2. What is the
primary risk of their implementation strategy?
Answer:
 The Foundation: Aiken’s Pyramid demonstrates that while Data Analytics is the
"golden function" that adds value, it requires a strong foundation of data modeling,
storage, and security to work.
 The Strategy Error: The HR manager is trying to reach Phase 4 (Advanced
Analytics) to boost productivity while intentionally weakening Phase 1
(Foundation) by choosing substandard security.
 The Risk: Without a reliable foundation, the "Advanced Analytics" results are at risk
of being compromised, and the organization faces significant Legal Liability—a
"weak signal" that could become a "strong signal" if an employee data breach occurs.
Question7: Case 1 mentions that customers may not be aware their interactions are
being "scrutinized." How does the "Building Transparency into AI" framework
address this?
Answer:
 Impact on Respect: Transparency expresses respect for people by preserving human
autonomy and preventing manipulation.
 Deception vs. Transparency: Using AI without disclosure is deceptive (similar to the
reservation AI example).
 Implementation: The company must communicate When an AI is being used
and Why it was chosen to allow customers to make informed choices
Question8: In Case 2, the HR manager chooses a "common open-source method" for
security to save costs. Using Aiken’s Pyramid, explain why this approach is
strategically flawed.
Answer:
 Aiken’s Pyramid Structure: The pyramid shows that organizations move through
four phases: Foundation (storage/security), Data Quality, Strategic Integration, and
finally Advanced Analytics.
36
  The Flaw: The HR manager is attempting to jump straight to Phase 4 (Advanced
Analytics) to predict turnover. However, they are intentionally weakening Phase 1
(The Foundation) by opting for subpar security.
 Consequence: Without a strong foundation, the "golden function" (analytics) is at
risk of failure or legal liability.

The Metaphor Study Checklist

Your notes use specific analogies to explain complex data concepts. These are key for
"Explain with an example" questions.
 The Kitchen Metaphor (Data Governance): * The Idea: You cannot deliver a good
meal just by having sharp knives (Technology). If the ingredients (Data) are wrong or
the menu (Strategy) isn't what the customer ordered, the meal fails.
o Exam Application (Case 1): In the marketing case, the "sharp knives" are the
advanced AI tools, but the "spoiled ingredients" are the biased social media
data.
 Renting an Apartment (Data Stewardship): * The Idea: A steward manages data
on behalf of others. Just as a tenant must maintain a rented apartment and not damage
it, a steward holds data "on trust" for the organization.
o Exam Application (Case 2): The HR manager is a "Decider" steward. By
choosing weak security, they are "damaging the apartment" (the employee
data) they were entrusted to protect.
 The Librarian Metaphor (Data Architecture): * The Idea: Data architects are like
librarians who create and maintain the organization’s knowledge about data structures
and vocabulary.
o Exam Application (Case 1): Architects must define exactly what "customer
sentiment" means (the vocabulary) so the AI doesn't produce "inaccurate or
unfair" results.

37