Week 1
Algorithms are encoded procedures for Algorithm
transforming unput data into a desired output, Set of rules to obtain
Input Output
based on specific calculations. the edpected output
from given input
Algorithmic power
(functions)
Four functions
- Prioritization making an ordered list
o Emphasize or bring attention to certain things at the expense of others (e.g. google page
rank)
- Classification picking a category
o Categorize a particular entity to given class by looking at any number of the entity's
features (e.g. inappropriate YT content)
- Association finding links
o Association decisions mark relationships between entities (e.g. Okcupid dating match)
- Filtering isolating what's important
o Including/excluding info according to various rules/criteria. Inputs to filtering algorithms
often take prioritizing, classification, or association decisions into account (FB TL)
Two broad categories
Rule-based algorithms
- Based on a set of rules/steps
- Typically IF THEN
- Pro: easy to follow
- Con: only applicable to specified conditions
Machine learning algorithms
- Algorithms that 'learn' by themselves (based on statistical models rather than deterministic rules)
- The algorithms are 'trained' based on a corpus of data from which they may 'learn' to make
certain kinds of decisions without human oversight
- Pro: flexible and amenable to adaptions
- Con: needs to be trained & black-box (at some point you won't understand the output of the
algorithm cause it trained itself)
- Most social media use machine learning algorithms big corpus of data
Recommender systems
- Recommender systems are algorithms that provide suggestions for content that is most likely of
interest to a particular user
o These algorithms that decide which content to display to who based on certain criteria
o Users hence receive distinct streams of online content
E.g. FB, Netflix, Spotify, YT, etc.
- Rationale: avoid choice overload, to maximize user relevance, and to increase work efficiency.
,Techniques (types of rec systems)
1. Content-based filtering: these algorithms learn to recommend items that are similar to the ones
that the user liked in the past (based on similarity of items)
2. Collaborative filtering: these algorithms suggest recommendations to the user based on items
that other users with similar tastes liked in the past
3. Hybrid filtering: combine features from both content-based and collaborative systems, and
usually with other additional elements (e.g. demographics) most common type.
Perceptions of algorithms appreciation (e.g. blind faith) vs aversion. May depend
on may factors
- Type of task
o E.g. mechanical, objective tasks (efficiency of showing relevant google content) vs
subjective (dating matches) See also lvl of subjectivity.
- Level of subjectivity in decisions
- Individual characteristics
- Etc. (more research needed)
Algorithmic persuasion
Algorithmic persuasion: 'any deliberate attempt by a persuader to influence the beliefs, attitudes and
behaviours of people through online communication that is mediated by algorithms'
Feedback loop of algorithmic persuasion
,Input
- First party data: data you yourself
- Second party data: Google, using data from a collaborative trusted party
- Third party data: external party, data brokerage, specialised in data.
- Explicit data: data we explicitly leave behind (writing our FB profile), we're aware we leave that
behind
- Implicit data: subconsciously: cookies, IP address, search history.
Algorithm
- Techniques (rule-based vs machine learning, different power structures class, prior, etc.)
- Objective of persuader: changing attitudes, opinions, feelings, behaviours, etc.
- Algorithmic bias
o Developers' bias: they have prejudice.
o Machine learning algorithms are trained on data sets, which can be flawed.
Persuasion attempt
- Context: commercial and non-commercial corporate comm, health comm, marketing, etc.
- Nature:
o Paid: e.g. sponsored. You paying for getting your ads shown by the algorithm
o Organic: if FB is using its own algorithm to filter your TL, that's organic (non-paid)
- Medium: smartphones, laptop, smart TV, public transport, etc.
- Modality: visual, audio, (audio)visual, conversational (Alexa)
Persuasion process
- Relevance: provide us most relevant content
- Reduction: compressing content is more persuasive easier to process
- Social norm: show us what peers find relevant
- Automation: we have bias, we over rely on technology. more vulnerable to alg pers.
- Reinforcement: they are reinforcing previous attitudes showing us content that fit within our
world view.
- Etc…
Persuasive effects
- Intended: desired, expected
- Unintended: undesired
o Manipulation
o Privacy issues
o Exploitation
o Vulnerable
, Week 2| Online advertising
Paper 1 Boerman: OBA literature overview
OBA: ''The practice of monitoring people's online behaviour and using the collected information to show
people individually targeted advertisements'' Boerman, 2017
OBA (online behavioural advertising) is a sub-group of personalized advertising.
Effects depend on advertiser- and consumer-controlled factors.
Advertiser-controlled
- Ad characteristics
o Level of personalization (extent)
Type of info used (web-browsing, clicks, basket)
Amount of info (only 1 type of data or combination)
o Accuracy !!!
- OBA transparency
o Privacy statements and informed consent
Shows what type of data they use and collect
have little effect because we don't read them (and hard to understand)
o Disclosure (increasing transparency through self-regulation)
hardly effective cause we barely recognize the symbol/know what it is)
Consumer controlled factors
- Knowledge and abilities
o Consumers have little knowledge about OBA and hold misconceptions
o Even less knowledge about legal protections (GDPR: EU regulation that protects privacy)
o Consumers do not seem to understand tools to protect online privacy
Algorithms are encoded procedures for Algorithm
transforming unput data into a desired output, Set of rules to obtain
Input Output
based on specific calculations. the edpected output
from given input
Algorithmic power
(functions)
Four functions
- Prioritization making an ordered list
o Emphasize or bring attention to certain things at the expense of others (e.g. google page
rank)
- Classification picking a category
o Categorize a particular entity to given class by looking at any number of the entity's
features (e.g. inappropriate YT content)
- Association finding links
o Association decisions mark relationships between entities (e.g. Okcupid dating match)
- Filtering isolating what's important
o Including/excluding info according to various rules/criteria. Inputs to filtering algorithms
often take prioritizing, classification, or association decisions into account (FB TL)
Two broad categories
Rule-based algorithms
- Based on a set of rules/steps
- Typically IF THEN
- Pro: easy to follow
- Con: only applicable to specified conditions
Machine learning algorithms
- Algorithms that 'learn' by themselves (based on statistical models rather than deterministic rules)
- The algorithms are 'trained' based on a corpus of data from which they may 'learn' to make
certain kinds of decisions without human oversight
- Pro: flexible and amenable to adaptions
- Con: needs to be trained & black-box (at some point you won't understand the output of the
algorithm cause it trained itself)
- Most social media use machine learning algorithms big corpus of data
Recommender systems
- Recommender systems are algorithms that provide suggestions for content that is most likely of
interest to a particular user
o These algorithms that decide which content to display to who based on certain criteria
o Users hence receive distinct streams of online content
E.g. FB, Netflix, Spotify, YT, etc.
- Rationale: avoid choice overload, to maximize user relevance, and to increase work efficiency.
,Techniques (types of rec systems)
1. Content-based filtering: these algorithms learn to recommend items that are similar to the ones
that the user liked in the past (based on similarity of items)
2. Collaborative filtering: these algorithms suggest recommendations to the user based on items
that other users with similar tastes liked in the past
3. Hybrid filtering: combine features from both content-based and collaborative systems, and
usually with other additional elements (e.g. demographics) most common type.
Perceptions of algorithms appreciation (e.g. blind faith) vs aversion. May depend
on may factors
- Type of task
o E.g. mechanical, objective tasks (efficiency of showing relevant google content) vs
subjective (dating matches) See also lvl of subjectivity.
- Level of subjectivity in decisions
- Individual characteristics
- Etc. (more research needed)
Algorithmic persuasion
Algorithmic persuasion: 'any deliberate attempt by a persuader to influence the beliefs, attitudes and
behaviours of people through online communication that is mediated by algorithms'
Feedback loop of algorithmic persuasion
,Input
- First party data: data you yourself
- Second party data: Google, using data from a collaborative trusted party
- Third party data: external party, data brokerage, specialised in data.
- Explicit data: data we explicitly leave behind (writing our FB profile), we're aware we leave that
behind
- Implicit data: subconsciously: cookies, IP address, search history.
Algorithm
- Techniques (rule-based vs machine learning, different power structures class, prior, etc.)
- Objective of persuader: changing attitudes, opinions, feelings, behaviours, etc.
- Algorithmic bias
o Developers' bias: they have prejudice.
o Machine learning algorithms are trained on data sets, which can be flawed.
Persuasion attempt
- Context: commercial and non-commercial corporate comm, health comm, marketing, etc.
- Nature:
o Paid: e.g. sponsored. You paying for getting your ads shown by the algorithm
o Organic: if FB is using its own algorithm to filter your TL, that's organic (non-paid)
- Medium: smartphones, laptop, smart TV, public transport, etc.
- Modality: visual, audio, (audio)visual, conversational (Alexa)
Persuasion process
- Relevance: provide us most relevant content
- Reduction: compressing content is more persuasive easier to process
- Social norm: show us what peers find relevant
- Automation: we have bias, we over rely on technology. more vulnerable to alg pers.
- Reinforcement: they are reinforcing previous attitudes showing us content that fit within our
world view.
- Etc…
Persuasive effects
- Intended: desired, expected
- Unintended: undesired
o Manipulation
o Privacy issues
o Exploitation
o Vulnerable
, Week 2| Online advertising
Paper 1 Boerman: OBA literature overview
OBA: ''The practice of monitoring people's online behaviour and using the collected information to show
people individually targeted advertisements'' Boerman, 2017
OBA (online behavioural advertising) is a sub-group of personalized advertising.
Effects depend on advertiser- and consumer-controlled factors.
Advertiser-controlled
- Ad characteristics
o Level of personalization (extent)
Type of info used (web-browsing, clicks, basket)
Amount of info (only 1 type of data or combination)
o Accuracy !!!
- OBA transparency
o Privacy statements and informed consent
Shows what type of data they use and collect
have little effect because we don't read them (and hard to understand)
o Disclosure (increasing transparency through self-regulation)
hardly effective cause we barely recognize the symbol/know what it is)
Consumer controlled factors
- Knowledge and abilities
o Consumers have little knowledge about OBA and hold misconceptions
o Even less knowledge about legal protections (GDPR: EU regulation that protects privacy)
o Consumers do not seem to understand tools to protect online privacy
