APG5457 Chap.4 Algorithms, Targeting and Opacity
Algorithms, Targeting and Opacity
Algorithmic systems shape exposure by ranking available material, but their operation is opaque to ordinary users and researchers. The visible feed is therefore evidence about ordering, not direct access to the model.
This chapter combines critical algorithm analysis with a rigorous auto-ethnographic method: capture a baseline, log ordinary actions, compare later screens, separate observation from inference, and connect patterns to scholarship without claiming certainty that the evidence cannot support.
Ranking is powerful even when personalisation is imperfect.
A system selects from available material, orders what appears first, repeats some items and leaves others difficult to find. The result structures attention and can affect political communication, advertising and cultural discovery. Avoid personifying the algorithm as if it had a stable intention.
Identify the firm, interface, signals and objective that may be involved, then mark the parts of the process that remain unavailable to inspection.
Fieldwork gives access to the lived relation rather than the source code. A baseline records categories, ordering, recurring material and prompts before a period of observation.
The action log captures searches, pauses, skips, follows, hides, purchases and the situation in which they occurred. Later screenshots make change comparable. One altered item proves little, but patterns across time can support a cautious interpretation. Popularity, experimentation and hidden signals remain plausible alternatives.
Reflexivity turns personal involvement into evidence.
Record how the interface shapes your next action as well as how your actions may supply signals. In the vignette, distinguish the observed screen, your emotional or practical response, the concept that explains the relation and the broader significance. Use calibrated verbs for hidden processes.
The aim is not maximal certainty; it is a transparent chain that lets the reader see what happened, how the interpretation was built and where its limits remain.
What this chapter covers
- 01
Ranking as an intervention in visibility
- 02
Targeting and continuous optimisation
- 03
Declared and behavioural signals
- 04
Filter-bubble claims and their limits
- 05
Algorithmic opacity
- 06
Baseline and later captures
- 07
Reflexive action logs
- 08
Vignettes, inference and significance
Interpret a changed feed carefully
- 2Describe the baseline and later ordering in concrete terms.
- 1Identify the logged behaviour that occurred between captures.
- 2Explain why the timing suggests a relation but does not prove a specific signal or model rule.
- 2Apply ranking and data-extraction concepts to the pattern.
- 1State the wider significance for categorisation and cultural exposure.
Key terms
- Algorithmic ranking
- The ordered presentation of available items according to computationally applied signals and objectives.
- Algorithmic opacity
- The difficulty of inspecting the inputs, rules and objectives that produce a ranked output.
- Behavioural trace
- Data generated through actions such as searching, pausing, watching, skipping, liking or purchasing.
- Baseline capture
- A documented starting view used to compare later platform content or ordering.
- Reflective vignette
- A situated account that connects a specific experience to concepts and broader cultural significance.
Algorithms, Targeting and Opacity FAQ
Can one recommendation reveal the algorithm?
No. One item may reflect popularity, testing, recent behaviour or another hidden signal. Use patterns across captures and distinguish visible output from inferred operation over time.
What belongs in an auto-ethnographic fieldnote?
Record time, visible categories and ordering, interface prompts, your actions, context and response. Keep observation separate from hypotheses and theoretical interpretation for careful later comparison.
How does a vignette become analytical?
Describe a concrete encounter, introduce the relevant evidence, apply a concept that explains the mechanism, acknowledge uncertainty and state the wider significance for platform culture or power.
Why is a baseline capture important?
A baseline records the initial categories, order and repeated material before the observation develops. It makes later change comparable and reduces dependence on memory or a retrospective story about the feed.
How should anomalies be handled in fieldwork?
Keep evidence that complicates the preferred explanation. Anomalies can reveal experimentation, hidden signals or weak patterns, and they help prevent one smooth causal story from replacing the uncertainty actually encountered.
What does reflexivity add to platform research?
Reflexivity explains how the researcher's searches, pauses, feelings and decisions participate in the relationship being studied. It also records how interface prompts shape the next action rather than imagining an external observer.
Which verbs are safest for an opaque system?
Use shows or places for visible interface evidence, suggests or is consistent with for patterns, and helps explain for theory-based interpretation. Avoid verbs that claim direct knowledge of hidden intentions or source code.
Assessment move
Keep two columns throughout fieldwork: direct observation and interpretation. Capture a baseline before deliberately altering behaviour, then log searches, pauses, follows, hides, skips and other ordinary actions with timestamps. Compare later screens by category, ordering, repetition and interface framing. Use calibrated verbs such as suggests, is consistent with and helps explain when the mechanism remains hidden.
For every screenshot, write what the reader should notice and how that feature supports or complicates the claim. Review personal and third-party details before submission.
