Course manual: https://canvas.uva.nl/courses/29268
W1: Videos
Video 1: What exactly is an algorithm?
- Algorithm: process/ set of rules to be followed in calculations or other problem-solving
operations, especially by a computer
- Set of instructions that enable a computer to put together different sources of
information and generate results (e.g., recipe to get a specific result)
- To deal with data (data can be anything!)
- Algorithms calculate based on features the sort of things that put
some things at the top of a list, and others at the bottom
- Process: Data goes in > goes through instructions > results come out
Coding vs. Algorithm
- Coding is the language of algorithms
Algorithm vs. humans
- No human error bc the computer goes through the instructions (and that's all they know
how to do)
- The human writing the code can make an error
Benefits of algorithms
- Speeding up decision-making
- Making whole processes efficient
- Maybe spotting things that humans have not spotted
W1: Readings
Defining concepts of the digital society
- Terminology shapes reality: phenomena addressed in research & terminology used for
research
- Terms & concepts: lenses on the complexity of reality that foreground some aspects
while neglecting others. Bear normative assumptions, install specific ways of
understanding new phenomena, create regulatory implications
- More frequently used = phenomena and its specific framing becomes increasingly
more self-evident and ordinary
- Hierarchisation of terms and ideas: choosing certain terms over others, giving
voice to a selection of authors, their respective disciplines and viewpoints
(problems: western perspectives, uneven mix of disciplinary positions, dominant
rep of certain auctorial subjectivities in terms of gender, race, ethnicity)
, - Algorithmic governance: builds on notation that technology allows for a specific
mode of governing society (alternative form of governing/ social order where
algorithms are applied to regulations)
- E.g. increasing amount of algorithmic systems (e.g. automated content
moderation) that increasingly rely on algorithmic governance
- Autonomous systems
- Transparency
- Smart technologies
- Platformization
- Implications: large-scale extraction of data = appropiation of social
resources with the general objective (mostly by Western companies) to
"dispossess"
- Filter bubbles
- Datafication: describes a cultural logic of quantification and monetisation of
human life through the digital information
- Implications for labour & establishment of new markets
- Basic rights of the self, autonomy and privacy are increasingly called into
question
Privacy in question
- Shift the possibilities & boundaries of human perception & action by creating
visbilities and forms of interaction that are no longer defined by physical presence
- E.g. personal pics potentially become accessible for a worldwide audience
- E.g. data is easy and cheap to store, become spermanent in digital records
- Data is argued to be used to identify behavioural patterns (not
same as personal data)
Zarouali, B., Boerman, S.C., Voorveld, H.A.M., & Van Noort, G. (2022). The Algorithmic Persuasion
Framework in online communication: Conceptualization and a future research agenda.
Internet Research
Goals of paper:
1. Define and conceptualize algorithmic persuasion
2. Proposal a fw that integrates different dynamic (e.g., data inout, algorithms, persuasion
attempts)
3. Develop future research agenda based on insights derivd from this fw
Algorithm: set of step-by-step instructions computers are programmed tofollow to accomplish
certain tasks
- Tranforming online platforms into codified environments that expose users to the
content that is likely to be most persuasive to them
,Aims of algorithms
- Persuade
- Increase value and capital
- Nudge behavor
- Change ppl's preferences
Concerns about algorithms
1. Fragmented public sphere
2. (Covrt) Highler likelihood of manipulation (voter/ consumers)
3. Increase in attitudinal polarization
4. More privacy infringements
5. Increase in user surveillance
6. Loss of user autonomy
7. Information asymmetry: imbalance in knwoeldgand decision-making
power favoring data processors over the user
8. Bias (bc coders unconsciously program their biases, e.g., prejudices,
stereotypes)
Algorithmic persuasion framework (APF) (in online communication)
- Algorithms have transformed online environments into persuasive architectures that
influence online choices of media users through unobtrusive and subtle processes
Algorithmic persuasion: "any deliberate attempt by a persuader to influence the beliefs, attitudes
and behaviors of people through online communication that is mediated by algorithms"
- Deliberate attempt: FOCUS ON INTENT OF PERSUADER! attempt is
purposefully initiated by the persuades (attempt to persuade does not have to be
recognized by receiver)
- Persuader: brand/ organization/ person
- Influence: change in people's beliefs, attitudes and behaviors (what ppl think, feel
and do) after expsorue to algorithm-mediated comm (varies from: v. weak - v.
strong; can be concious/ unconscious level; ST/LT; intended/unintended)
- Online communication: transmission of algorithm-mediated comm from senters to
receivers in online environments (always online comm activity driven by
algorithms)
- Algorithm-mediated: online comm must be mediated by algorithms that
automatically decide which content to select and present to which users based on
large corpus of input data
- Dynamic process
- 5 components that are central to
algorithmic persuasion
, (circular;feedbsack loops, all are cause & effect of each other): input, algorithm,
persuasion, attempt, persuasion process, persuasion effects
Component 1: Input: Data
- Involves all data that is ussed in algorithmic persuasion, "customer data"
- Prior to algorithm-mediated comm can be provided to online users, data has to be
collected and readied for the algorithm
- First-party data: data owned/ collected by the sender (soruce of persuasion
attempt), you collected it, you own it
- Collected through: cookies, dta about online purchases, data
entered when receiver becomes a member of/ donator to a pol
party, data disclosed during registation of a device
- Second-party data: data that can be used for algorithmic persuasion bc
they are owned (BOUGHT) by a collaborative party
- E..g, buy media space in an automated what based on real-time
bidding (buy data from Google)
- Third-party data: data collected by companies that are not directly
involved in the primary process, persuaders purchase data from data
brokers that specilize in collecting and combining data
- Explicit data: (WITTINGLY) data wittingly (with knowledge/ deliberately)
disclosed by users in online environments
- Implicit data: (UNWITTINGLY) compilation of and/or inferences from data
about users colleced w/out their awareness (e.g., surfing history, preferences, IP
address)
Component 2: Algorithm: Techniques, persuader, objectives, biases
- Algorithms: encoded procedures for transforming input dasta into desired outputs, based
on specified calculations. Data is convreed into "algorithm friendly" format
- (Simple) rule-based algorithms: based on a set of rules/steps
- Typically "IF-THEN" statements → if person likes x content then give y
as result
- Pos: quick to formulate & easy to follow bc it reads as plain text)
- Neg: only applicable to condition in which they are formulated, only for
specific platforms bc different platforms have different "if"s
- (Complex) machine learning algorithms: algorithms that "learn" by themselves
(based on statistical models rather than deterministic rules, trianed with large
corpus of data)
- "Trained" and therfreore "learn"to make decisions w/out human oversight
