The costs of the attention machine: surveillance, manipulation, and the questions to ask before we leave.
Unit D: Why does it matter?
By the end of today
What you will walk out knowing
01
Evaluate surveillance and manipulation
Distinguish surveillance (watching without consent) from manipulation (shaping behaviour without transparency), and identify where each appears in the systems we have studied.
Reveal
02
Apply the perverted public interest concept
Use Carlson's framework to explain how optimising for engagement metrics can corrupt the public interest mission of journalism, and identify equivalent dynamics in OTT, games, and creator platforms.
Reveal
03
Assess engagement harm and algorithmic risk
Evaluate the evidence on algorithmic amplification of outrage, youth mental health effects, addictive design, and political polarisation, separating empirical claims from moral panics.
Reveal
04
Evaluate ethical frameworks for audience research
Apply a basic ethical test to the audience practices examined in this course: informed consent, benefit to the audience, harm minimisation, and accountability.
Reveal
Every system we have studied this term was built by someone, for someone. Who benefits when an audience is monitored, segmented, and optimised? And who bears the cost?
The full picture
What the attention economy costs
Privacy: audience monitoring collects personal behavioural data at scale, largely without meaningful informed consent. The asymmetry between what platforms know and what audiences know is structural, not accidental.
Autonomy: recommendation systems trained on past behaviour narrow future experience. An audience that only sees what algorithms predict it will like is being shaped, not served.
Public interest: the "perverted public interest" occurs when the desire to serve audiences is conflated with the interest to attract and retain them through manipulation, outrage, and clickbait.
Wellbeing: algorithms optimised for engagement amplify emotionally triggering content. The aggregate effect on political polarisation, trust in institutions, and youth mental health is contested but real.
Accountability: the systems that govern audience experience are largely proprietary, unaudited, and invisible to the audiences they govern.
Carlson (2018) · measurable journalism
The perverted public interest
"A desire to serve the public interest is conflated with an interest to manipulate for business and marketing purposes."
The original tension: journalism has always distinguished between stories that interest the public and stories that serve the public interest. The first is measurable; the second is harder to quantify.
The metric pressure: when newsrooms optimise for clicks, shares, and time on page, they create structural incentives toward outrage, conflict, and emotional arousal over calm, important coverage.
The practical manifestation: headline tweaking for algorithmic performance, "clickbaiting practices" that mislead audiences about content, deprioritising investigations in favour of viral stories.
The "slavery to metrics": "stories with public interest value are sacrificed for those with high viral potential." The growing tendency to follow audience preferences means journalism loses its traditional role of setting the agenda.
In Bangladesh: the same Prothom Alo that broke factory safety investigations also optimises Facebook headlines for engagement. Both are true, simultaneously, of the same organisation.
The engagement critique
When algorithms hurt audiences
01
Emotional amplification
Algorithms reward content that triggers strong emotional responses: outrage, fear, indignation, tribal solidarity. This content outperforms calm, balanced content. The aggregate effect tilts public discourse toward emotional extremes.
Reveal
02
Political polarisation
Research on algorithmic curation and political polarisation is contested. Algorithms do amplify polarised content, making it more visible than its actual prevalence warrants. But whether algorithms cause polarisation, or simply reflect existing divisions, remains an active empirical debate.
Reveal
03
Youth mental health
Former platform employees and independent researchers have documented correlations between heavy Instagram use and body image anxiety in young women. The platforms' own internal research acknowledged these effects before they became public. Correlation is not causation, but the asymmetry of information is troubling.
Reveal
04
Addictive design
Free-to-play games, short-form video, and social media all use variable reward schedules borrowed from gambling psychology. The design is intentional. Whether it crosses from engaging to addictive depends on contested definitions, but the design intention is engagement maximisation, not user wellbeing.
Reveal
Local context
Ethical dimensions specific to Bangladesh
The Digital Security Act: Bangladesh's regulatory environment for digital speech is complex and contested. Provisions have been used to prosecute social media users for content the government deems objectionable, with substantial implications for press freedom and audience safety in expression.
Platform indifference: global platforms apply moderation policies with limited adaptation to Bangladeshi contexts. Content that is removed quickly in high-revenue markets may persist longer in lower-revenue markets.
Misinformation at scale: Facebook's dominance as a news distribution platform in Bangladesh means that misinformation spreads through the same channel as trusted journalism. Audiences often cannot distinguish between them in a feed context.
Data sovereignty: most data generated by Bangladeshi audiences is stored and processed by US-headquartered firms, under US legal frameworks, in ways that are not subject to meaningful Bangladeshi regulatory oversight.
The measurement gap: Bangladesh's nascent audience measurement infrastructure means that many ethical abuses in targeting and content manipulation are harder to detect and document than in markets with more established independent measurement.
Critical framework
Surveillance, manipulation, and the audience
01
Surveillance
Watching without meaningful consent. The audience does not know the extent of what is collected, cannot access it, and cannot meaningfully opt out without abandoning services that are near-essential. Structural, not incidental.
Reveal
02
Manipulation
Shaping behaviour through design rather than persuasion. Variable reward schedules, algorithmic amplification of outrage, personalised pricing, dark patterns in subscription cancellation flows. The intent is to produce behavioural outcomes the user would not choose if fully aware.
Reveal
03
The information asymmetry
Platforms know an extraordinary amount about users. Users know almost nothing about how platforms use that information. This asymmetry is not accidental. The long-term sustainability of the relationship depends on audiences trusting that the asymmetry will not be abused.
Reveal
04
The accountability gap
The systems that govern audience experience are proprietary, frequently changed, and almost completely unaudited by external parties. When harm occurs, it is nearly impossible to establish causation and nearly impossible to hold any specific actor accountable.
Reveal
Regulatory landscape
How regulators are responding
GDPR (EU, 2018): individuals must have meaningful control over data collection. Firms must obtain genuine consent, use data only for consented purposes, and delete it on request. Has effectively set a global standard.
App Tracking Transparency (Apple, 2021): requires users to opt in to cross-app tracking. Estimated to have reduced Meta's revenue by approximately ten billion dollars in its first year. A private regulatory intervention with public market consequences.
Digital Services Act (EU, 2024): imposes obligations on large platforms regarding content moderation, transparency, and user appeals. Requires disclosure of algorithmic recommendation logic to regulators.
Cookie deprecation: major browsers have deprecated or are deprecating third-party cookies, dismantling the infrastructure of behavioural advertising and forcing a shift to first-party data strategies.
Bangladesh trajectory: the country's Personal Data Protection Act has been in draft for several years. Firms operating in Bangladesh should anticipate a regulatory environment that will move closer to GDPR-adjacent frameworks within the next few years.
Stance spectrum
Stance spectrum
Engagement optimisation is a form of manipulation.
Strongly agreeStrongly disagree
Drag the slider to take a position
Branching scenario · Ethics desk
Branching scenario
Three decisions, one platform
You are head of editorial ethics at a major Bangladeshi news platform. Three proposals arrive on your desk the same morning. Each promises to increase engagement. Each raises a different ethical concern. You can only examine one at a time.
Proposal A: Reject. Headline bait-and-switch is the textbook "perverted public interest" move. It uses the audience's attention to generate a click the content does not earn. It erodes trust over time as audiences notice the gap between headline and content. The short-term engagement gain is real; the long-term trust cost is larger.
Proposal B: Revise. Personalisation is not inherently unethical, but a system that systematically removes audiences from uncomfortable topics is an autonomy problem. Revision should include: user control over the algorithm, exposure to a minimum diversity threshold, and transparency about what the system is doing and why.
Proposal C: Approve. This is the Hearken model applied: reciprocal engagement that gives the audience genuine input into editorial decisions. The commitment to publish regardless of traffic potential is the ethical core. It inverts the perverted public interest dynamic by making editorial choices accountable to the audience, not the metric.
Sorting game
Sorting game
Does this practice pass an ethical test?
Apply four criteria: informed consent, audience benefit, harm minimisation, accountability. Sort each practice.
Practices to sort
Knowledge check
Knowledge check
Carlson and colleagues describe a "perverted public interest" in digital journalism. What does this term mean?
D. The perverted public interest does not mean deliberately false or dishonest journalism. It describes a structural drift: the desire to serve the public interest is real, but the pressure to optimise for engagement metrics pulls editorial decisions toward what attracts and retains audiences, which is not always what serves them.
The evidence clearly establishes that social media algorithms cause political polarisation, and this is the settled scientific consensus.False. The relationship between algorithmic curation and political polarisation is contested, not settled. Research shows that algorithms amplify polarised content and make it more visible than its actual prevalence warrants. But whether algorithms cause polarisation, or simply reflect existing divisions, remains an active empirical debate.
Apple's App Tracking Transparency framework, which requires users to opt in to cross-app tracking, was estimated to have reduced Meta's revenue by approximately ten billion dollars in its first year.True. ATT's opt-in requirement dramatically reduced the data available for targeted advertising on iOS devices. This illustrates how a single private regulatory intervention can have major commercial consequences for the entire digital advertising ecosystem.
Your turn · 15–20 min
Activity: The Ethical Audit
Steps 1–2
Choose a practice and apply the four criteria
Pick one specific thing a Bangladeshi organisation does to its audience. Test it against the four criteria: informed consent, audience benefit, harm minimisation, and accountability. Be concrete in each cell.
Step 3
Surveillance or manipulation?
Classify the practice using today's distinction. Does it involve watching the audience without meaningful awareness? Does it shape behaviour without transparency? It may be both. Explain which dimension is stronger and why.
Steps 4–5
Perverted public interest and one recommendation
Does this practice corrupt the organisation's public interest mission? Then write one concrete change that would make the practice more ethical. Not "be more transparent" but what, how, for whom.
The three questions
What this course set out to answer
Q1Who is the digital audience? Datafied, active, participatory, segmented, algorithmic. Not a mass; not a monolith. A shifting, platform-mediated, commercially constructed category that academic and practitioner traditions have contested since the first ratings diary.
Q2How do digital audiences behave? With attention governed by genre expectations and taste; with participation ranging from clicking to co-creating; on journeys shaped by discovery algorithms and parasocial bonds; in ways that produce commercial value they rarely share in.
Q3Why research and monitor them? Because attention is scarce and valuable; because audiences are both the receiver of content and the commodity sold to advertisers; because understanding who your audience is and what they need is the prerequisite for serving them, and the prerequisite for exploiting them.
The answer to Q3 depends entirely on what you do with what Q1 and Q2 taught you. Your ethical audit today is the first time you apply all three answers to a single practice.
Today in five lines
Recap
01The attention economy has costs: privacy loss, autonomy erosion, public interest corruption, wellbeing harm, and accountability gaps.
02The perverted public interest describes how engagement optimisation structurally pulls journalism away from its public mission without anyone deciding to corrupt it.
03Surveillance (watching without consent) and manipulation (shaping behaviour without transparency) are distinct ethical problems requiring distinct responses.
04Regulatory responses (GDPR, ATT, DSA) are reshaping the infrastructure of the attention economy, but the pace of regulation consistently lags the pace of platform innovation.
05The ethical test: informed consent, audience benefit, harm minimisation, accountability. Apply it to every system we have studied, including the ones you will build.
Digital Audience · Lecture 15
Know your audience. Serve them.
This week’s reading
Reading list
PrimaryNightingale (ed.) — Ch. 8, “The Implied Audience of Communications Policy Making” (Livingstone & Lunt)