Wissensmedien
Vorlesung Wissensmedien Jürgen Buder
Vorlesung Wissensmedien Jürgen Buder
-
- 1 / 356
-
Cartes-fiches
What is the output view?
Humanlikeness if they can do the same as humans
What is the response view?
Humanlikeness if people don’t treat computers differently from humans
Do LLMs exhibit cognitive dissonance and theory of mind?
No, those claims have been refuted
What are LLMs lacking?
- World knowledge
- Functional mastery of language
Emotion, Motivation and agency
What is the conclusion of the structural view?
Computers are not humanlike
What are robots definitely worse in than humans?
Motor performance
What are similarities in the psychology of AI and humans?
- Humanlike cognitive performance
- Wisdom of the crowd effects
- Same biases and stereotypes
Racism
What is the conclusion of the output view?
Computers are not humanlike because they are and certainly will be better in most things than us
What is Anthropomorphism?
Seeing something human in non-human entities
Which two types of anthropomorphism can we differentiate?
- Looking human-like
Acting human-like
What does the mind perception approach postulate?
Entities are evaluated on two dimensions:
- Agency
Experience
How are robots commonly rated on agency and experience?
- High on agency
Low on experience
Which four facets does the building blocks of anthropomorphism theory contain?
- Cognitive ability
- Autonomy
- Subjectivity
Cooperation
Which Dimension of the building blocks theory does consist of subjectivity and cooperation? What does it mean?
- Sociality
- Like experience
Which Dimension of the building blocks theory does consist of cognitive ability and autonomy? What does it mean?
- Intelligence
Like Agency
What is the first prediction of the BBA theory?
The presence of each building block increases anthropomorphism
What is the second prediction of the BBA theory?
If we don’t know that we interact with a computer, people behave in line with CASA
What is the third prediction of the BBA theory?
If we do know that we interact with a computer, people behave in line with mind perception
What happens when a source is disclosed? (Mind perception)
- We ascribe high cognitive ability and cooperation to computers and expect them to behave in a competent and friendly manner
We ascribe low to medium subjectivity and autonomy to computers and expect them not to show subjectivity and autonomy
Who do people take as an advisor for cognitive tasks, whom for assessing emotions?
- Computer for cognitive ability
Humans for subjectivity tasks
Study on feeling heard, who induces the highest “feeling heard effect”?
- AI responses with human label
Gets diminished when labeled as AI
What are the four assumptions of the BBA?
- Computers are tools to help us accomplish tasks
- Tools require reliability and certainty
- Cognitive load and cooperation are beneficial
Autonomy and subjectivity are “scary”
What are the fourth and fifth prediction of the BBA?
- Too little cognitive ability or cooperation = negative evaluation of robot
Too much autonomy or subjectivity = negative evaluation
What does higher anthropomorphism lead to? Does it depend on something?
Higher anthropomorphism does not automatically lead to more liking
• This depends on source disclosure
• Without source disclosure (Turing studies), we see no difference between humans and (capable) computers
What is the conclusion of the response view?
- Without source disclosure computers are humanlike
With source disclosure not
When is cognitive dissonance exhibited (examples)?
- When the world doesn’t end
When people disagree on social media
How can we justify a made choice?
Finding post hoc arguments for the chosen option
Explain what forced-compliance experiments are
- People were supposed to lie about task they did (was boring)
- People with big reward found the reward to be a good excuse for participation
Small reward forced them to believe they had intrinsic motivation (task was sooo great!)
What is thought to happen when people have dissonance?
- Reduction of dissonance
Which options are there to reduce dissonance? (4)
- Conversion (give up belief)
- Derogation (reduce importance)
- Lying to oneself
Adding consonant cognitions
What does adding consonant cognitions eventually lead to?
- Selective exposure
Only by support of likeminded individuals (e.g. filter bubbles online)
To which type of information do people tend to expose themselves?
- Congenial information (confirmation/congeniality bias)
Smoking example
What is the minimal effects paradigm?
- Communication science
- Only already believing people are reached
There is a likewise effect in sociology (homophily)
What is the average effect size for the congeniality bias?
D = .36
What increases congeniality bias?
- Close-mindedness
Value-laden topics
What does selective exposure matter in terms of (US-) politics?
- Can explain polarization
Republicans and democrats choose different medias
What happened when people had to pick sources for different topics?
People chose sources than align with political view even for not political topics
In which context did the term “echo chambers” emerge?
- Creation of the internet
Cass Sunstein
What found the twitter study?
Retweets are typically sent to like-minded individuals (only for political topics)
What differentiates echo chamber and filter bubbles?
Filter bubbles mean the presentation of congenial information by algorithms
