Algorithmic Persuasion
Lecture 1
Introduction to Algorithms
Algorithms: encoded procedures for transforming input data into a desired output,
based on specified calculations (Gillespie, 2014)
Input Algorithm (set of rules to obtain the expected output given the
input) Output
Algorithmic Power
1. Prioritization: making an ordered list; emphasise or bring attention to certain things
at the expense of others.
Ex: Google page ranks; what comes first depends on the person searching.
2. Classification: picking a category; categorize a particular entity to given class by
looking at any number of that entity’s features
Ex: inappropriate YouTube content
3. Association: finding links; association decisions mark relationships between entities
Ex: OKCupid dating match.
4. Filtering: isolating what’s important; including or excluding information according to
various rules of criteria. Inputs to filtering algorithms often take prioritizing,
classification or association decisions into account.
Ex: Facebook news feed.
Types of algorithms
Rule-based algorithms
Simple, based on specific steps or rules.
IF THEN statements IF ‘condition’ THEN ‘result’
+: quick easy to follow
-: but only applicable to the specified conditions, time consuming
Machine learning algorithms
Algorithms that learn by themselves (based on statistical models rather than rules)
These algorithms are “trained” based on a large set of data from which they try to
learn/find patterns to make certain decision without human oversight
+: flexible and amenable to adaptation.
-: need to be trained and black-box (not knowing why the algorithm made
the decision).
In data-intensive online environments, ML algorithms have become the standard
, Logic: train the algorithm on a sample of data and then it can be used for making
predictions about other data
Facebook, Amazon, Netflix, etc. use ML algorithms
They have loads of data, thus the machine has a lot to learn
A few hundred lines of code can easily generate a model consisting of
millions of lines.
Example: Facebook’s DeepFace algorithm
Facial recognition system
It identifies human faces in digital images
Trained on a large “identity labelled dataset” of four million facial images
97% accuracy
Zuboff: these systems can be sold to businesses and authoritarian regimes.
Level of Automation In Algorithms
Recommender Systems
Recommender Systems are algorithms that provide suggestions for content that is
most likely of interest to a particular user.
These algorithms that decide which content to display to whom based on
certain criteria
Users are thus receiving distinct streams of online content
Facebook news feed, Netflix movie suggestions, songs on Spotify,
videos on YouTube, products on Amazon, etc.
Rationale: avoid choice overload, to maximise user relevance, and to increase work
efficiency.
, 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)
Collaborative filtering: these algorithms suggests recommendations to the user
based on items that other users with similar tastes liked in the past
Hybrid filtering: these algorithms combine features from both content and
collaborative systems, and usually with other additional elements such as
demographics.
Netflix’s recommender system.
Views On Algorithms
Algorithmic Appreciation
People rely more on advice from algorithms than from other people, despite
blindness to algorithm’s process (“black-box”).
Automation bias: humans tend to over-rely on automation (blind faith in
information from computers).
Information from automation > information from humans
Humans are imperfect, whereas computers are objective, rational, neutral,
reliable, etc.
Ex: GPS systems, automatic pilot, spelling checker, etc.
Algorithmic Aversion
Tendency to prefer human judgements over algorithmic decisions ,even when the
human decisions are clearly inferior
Less tolerance for errors from algorithms than from humans
Ex: GPS leads you into a traffic jam, you’d over-react.
People are averse because they don’t understand the algorithmic process
They think human decisions are easier to understand, but this is subjective!
Algorithmic anxiety: lack of control and uncertainty over algorithms creates anxiety
among Airbnb hosts.
Aversion or appreciation depend on many other factors:
Type of task
Level of subjectivity in decisions
Individual characteristics
Etc. (more research is needed)
Also, the choice is not always “human vs algorithm”
Lecture 1
Introduction to Algorithms
Algorithms: encoded procedures for transforming input data into a desired output,
based on specified calculations (Gillespie, 2014)
Input Algorithm (set of rules to obtain the expected output given the
input) Output
Algorithmic Power
1. Prioritization: making an ordered list; emphasise or bring attention to certain things
at the expense of others.
Ex: Google page ranks; what comes first depends on the person searching.
2. Classification: picking a category; categorize a particular entity to given class by
looking at any number of that entity’s features
Ex: inappropriate YouTube content
3. Association: finding links; association decisions mark relationships between entities
Ex: OKCupid dating match.
4. Filtering: isolating what’s important; including or excluding information according to
various rules of criteria. Inputs to filtering algorithms often take prioritizing,
classification or association decisions into account.
Ex: Facebook news feed.
Types of algorithms
Rule-based algorithms
Simple, based on specific steps or rules.
IF THEN statements IF ‘condition’ THEN ‘result’
+: quick easy to follow
-: but only applicable to the specified conditions, time consuming
Machine learning algorithms
Algorithms that learn by themselves (based on statistical models rather than rules)
These algorithms are “trained” based on a large set of data from which they try to
learn/find patterns to make certain decision without human oversight
+: flexible and amenable to adaptation.
-: need to be trained and black-box (not knowing why the algorithm made
the decision).
In data-intensive online environments, ML algorithms have become the standard
, Logic: train the algorithm on a sample of data and then it can be used for making
predictions about other data
Facebook, Amazon, Netflix, etc. use ML algorithms
They have loads of data, thus the machine has a lot to learn
A few hundred lines of code can easily generate a model consisting of
millions of lines.
Example: Facebook’s DeepFace algorithm
Facial recognition system
It identifies human faces in digital images
Trained on a large “identity labelled dataset” of four million facial images
97% accuracy
Zuboff: these systems can be sold to businesses and authoritarian regimes.
Level of Automation In Algorithms
Recommender Systems
Recommender Systems are algorithms that provide suggestions for content that is
most likely of interest to a particular user.
These algorithms that decide which content to display to whom based on
certain criteria
Users are thus receiving distinct streams of online content
Facebook news feed, Netflix movie suggestions, songs on Spotify,
videos on YouTube, products on Amazon, etc.
Rationale: avoid choice overload, to maximise user relevance, and to increase work
efficiency.
, 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)
Collaborative filtering: these algorithms suggests recommendations to the user
based on items that other users with similar tastes liked in the past
Hybrid filtering: these algorithms combine features from both content and
collaborative systems, and usually with other additional elements such as
demographics.
Netflix’s recommender system.
Views On Algorithms
Algorithmic Appreciation
People rely more on advice from algorithms than from other people, despite
blindness to algorithm’s process (“black-box”).
Automation bias: humans tend to over-rely on automation (blind faith in
information from computers).
Information from automation > information from humans
Humans are imperfect, whereas computers are objective, rational, neutral,
reliable, etc.
Ex: GPS systems, automatic pilot, spelling checker, etc.
Algorithmic Aversion
Tendency to prefer human judgements over algorithmic decisions ,even when the
human decisions are clearly inferior
Less tolerance for errors from algorithms than from humans
Ex: GPS leads you into a traffic jam, you’d over-react.
People are averse because they don’t understand the algorithmic process
They think human decisions are easier to understand, but this is subjective!
Algorithmic anxiety: lack of control and uncertainty over algorithms creates anxiety
among Airbnb hosts.
Aversion or appreciation depend on many other factors:
Type of task
Level of subjectivity in decisions
Individual characteristics
Etc. (more research is needed)
Also, the choice is not always “human vs algorithm”
