Lecture Notes Language Technology & Society
Table of Contents
Lecture Notes Language Technology & Society......................................................................................1
Lecture 1; Introduction, ‘smart’ technology, ethics and AI.................................................................3
Hypes and panics............................................................................................................................3
Historical context............................................................................................................................3
Technology is never neutral............................................................................................................4
Examples of language technology..................................................................................................4
Questions to ask.............................................................................................................................4
Example - Machine Translation......................................................................................................5
Direct and Indirect Stakeholders....................................................................................................6
Questions to ask (about new technologies):...................................................................................6
Lecture 2; How does ‘smart’ technology work? (AI/Machine learning and Evaluation metrics)........7
Modularity......................................................................................................................................7
Algorithms......................................................................................................................................7
Machine learning............................................................................................................................8
Evaluation metrics........................................................................................................................11
How does machine learning relate to algorithms?.......................................................................13
Lecture 3; Language Technology: NLP, NLG, Chatbots, Speech technology and virtual assistants...14
What is NLP?.................................................................................................................................14
Part 1: Language Understanding...................................................................................................15
Summary Part 1:...........................................................................................................................19
Part 2: Language production.........................................................................................................19
Summary Part 2:...........................................................................................................................22
Chatbots and dialog systems........................................................................................................22
Personal Assistants.......................................................................................................................23
Summary – Personal Assistants....................................................................................................26
Lecture 4; Hype and Technology adoption.......................................................................................27
Technology has a history..............................................................................................................27
Gartner Hype Cycle.......................................................................................................................27
Technology Life Cycle (Anderson & Tushman, 1990)....................................................................30
Models of Adoption......................................................................................................................31
Is a hype worth it?........................................................................................................................32
,Lecture 5; Panic and Resistance........................................................................................................33
A history of social concerns about technology.............................................................................33
Technology worry/panic is a recurring phenomenon...................................................................34
The Sisyphean Cycle of Technology Panics (Amy Orben 2020).....................................................35
Conceptualizing technology influence..........................................................................................38
A Typology of ‘Technology effects’...............................................................................................38
Summary – The effects of technology..........................................................................................39
Lecture 6; A theoretical vocabulary for understanding the impact and role of (language) technology
..........................................................................................................................................................41
Nuance is key................................................................................................................................41
Sociological level of Analysis (macro-level)...................................................................................41
Psychological level of Analysis......................................................................................................45
Summary......................................................................................................................................48
Lecture 1; Introduction, ‘smart’ technology, ethics and AI.................................................................2
Lecture 2; How does ‘smart’ technology work? (AI/Machine learning and Evaluation metrics)........6
Lecture 3; Language Technology: NLP, NLG, Chatbots, Speech technology and virtual assistants...13
Lecture 4; Hype and Technology adoption.......................................................................................26
Lecture 5; Panic and Resistance.......................................................................................................32
Lecture 6; A theoretical vocabulary for understanding the impact and role of (language) technology
.........................................................................................................................................................40
,Lecture 1; Introduction, ‘smart’ technology, ethics and AI
The media always been covering a lot of articles regarding new technologies. This leads people to be
aware of these new technologies and their abilities, and strengthens both (I) excitement about what
these technologies can do, and (II) fears about what they will do.
Hypes and panics
Hype: when people are overly excited about some new technology.
Example 1:
Self-driving cars; in 2015, Elon Musk said self-driving cars that could drive “anywhere” would be here
within two or three years. did not happen yet.
Example 2:
Artificial intelligence; in 2015, Mark Zuckerberg said “one of Facebook’s goals for the next five to 10
years is to basically get better than human level at all of the primary human sense: vision, hearing,
language, general cognition.” also not very realistic, a bit over-the-top.
Panic: when people are overly worried about some new technology.
Example 1:
Bicycles: in 1894 the New York Times proclaimed: “there is not the slightest doubt that bicycle riding,
if persisted in, leads to weakness of mind, general lunacy, and homicidal mania.” also did not
happen.
Example 2:
Filter bubbles: in 2011, Eli Pariser introduced the concept of filter bubbles: people becoming isolated
from other opinions, because news & social media algorithms feed them only items that they like and
agree with. Is this a real issue? Or another unwarranted panic? Need to research that.
Hypes and panics might form a feedback loop:
Questions for this course:
- How can we find the right balance between hype and panic?
- How can we assess the true effects and capabilities of smart technology?
Historical context
Short history about AI:
1950: Alan Turing publishes his “Computing Machinery and Intelligence”, proposing what is
know as The Turing Test.
1956: The Dartmouth Summer Workshop on Artificial Intelligence.
1966: Joseph Weizenbaum publishes ELIZA, the first chatbot.
Until 1970s: optimism!
1973: Lighthill report (skeptic about AI), start of first AI winter (period in which people where
negative about the idea of AI, lasted about 8-10 years).
Early 1980s: careful optimism again.
Late 1980s, early 1990s: second AI winter.
, 1990s, 2000s: rise of Machine Learning.
2010s: rise of Deep Learning.
Mid-2010s – Now: optimism … but until when?
Technology is never neutral
Technologies are never neutral. They can for example benefit one group in the society, but another
group of people in the society is not able to use it.
Example 1: Hearing aids
Pro: better hearing, connecting with others.
Con: hearing unwanted sounds, blamed for disability.
Without technology: patient.
With technology: ‘unwilling to do anything about it’.
Example 2: Soap dispenser (with sensor)
Pro: Improved hygiene.
Con: Darker skin colors are not recognized by some of these sensors.
So, the way that technology is designed impacts who gets to use it. Therefore, very important to test
technologies on multiple kinds of people, situations, etc.
Example 3: Image recognition systems
Pro:
Improved searchability;
Automation of tedious tasks (e.g., going through
all your pictures).
Con:
Could lead to an invasion of privacy;
Labels impose/encode a particular world view:
which objects are relevant?
o influences follow-up work: people
often study/use what is easy to
study/use (e.g., focus on Twitter data,
because it is easy to get).
Examples of language technology
1. Chatbots;
2. Voice assistants;
3. Grammar check, autocorrect;
4. Search engines/recommender systems (e.g., Google, YouTube, Spotify);
5. Sentiment analysis;
6. Spam filtering, content moderation.
Questions to ask
- What happens if it works well?
- What happens if it doesn’t?
- Who is this system likely to be successful for?
Table of Contents
Lecture Notes Language Technology & Society......................................................................................1
Lecture 1; Introduction, ‘smart’ technology, ethics and AI.................................................................3
Hypes and panics............................................................................................................................3
Historical context............................................................................................................................3
Technology is never neutral............................................................................................................4
Examples of language technology..................................................................................................4
Questions to ask.............................................................................................................................4
Example - Machine Translation......................................................................................................5
Direct and Indirect Stakeholders....................................................................................................6
Questions to ask (about new technologies):...................................................................................6
Lecture 2; How does ‘smart’ technology work? (AI/Machine learning and Evaluation metrics)........7
Modularity......................................................................................................................................7
Algorithms......................................................................................................................................7
Machine learning............................................................................................................................8
Evaluation metrics........................................................................................................................11
How does machine learning relate to algorithms?.......................................................................13
Lecture 3; Language Technology: NLP, NLG, Chatbots, Speech technology and virtual assistants...14
What is NLP?.................................................................................................................................14
Part 1: Language Understanding...................................................................................................15
Summary Part 1:...........................................................................................................................19
Part 2: Language production.........................................................................................................19
Summary Part 2:...........................................................................................................................22
Chatbots and dialog systems........................................................................................................22
Personal Assistants.......................................................................................................................23
Summary – Personal Assistants....................................................................................................26
Lecture 4; Hype and Technology adoption.......................................................................................27
Technology has a history..............................................................................................................27
Gartner Hype Cycle.......................................................................................................................27
Technology Life Cycle (Anderson & Tushman, 1990)....................................................................30
Models of Adoption......................................................................................................................31
Is a hype worth it?........................................................................................................................32
,Lecture 5; Panic and Resistance........................................................................................................33
A history of social concerns about technology.............................................................................33
Technology worry/panic is a recurring phenomenon...................................................................34
The Sisyphean Cycle of Technology Panics (Amy Orben 2020).....................................................35
Conceptualizing technology influence..........................................................................................38
A Typology of ‘Technology effects’...............................................................................................38
Summary – The effects of technology..........................................................................................39
Lecture 6; A theoretical vocabulary for understanding the impact and role of (language) technology
..........................................................................................................................................................41
Nuance is key................................................................................................................................41
Sociological level of Analysis (macro-level)...................................................................................41
Psychological level of Analysis......................................................................................................45
Summary......................................................................................................................................48
Lecture 1; Introduction, ‘smart’ technology, ethics and AI.................................................................2
Lecture 2; How does ‘smart’ technology work? (AI/Machine learning and Evaluation metrics)........6
Lecture 3; Language Technology: NLP, NLG, Chatbots, Speech technology and virtual assistants...13
Lecture 4; Hype and Technology adoption.......................................................................................26
Lecture 5; Panic and Resistance.......................................................................................................32
Lecture 6; A theoretical vocabulary for understanding the impact and role of (language) technology
.........................................................................................................................................................40
,Lecture 1; Introduction, ‘smart’ technology, ethics and AI
The media always been covering a lot of articles regarding new technologies. This leads people to be
aware of these new technologies and their abilities, and strengthens both (I) excitement about what
these technologies can do, and (II) fears about what they will do.
Hypes and panics
Hype: when people are overly excited about some new technology.
Example 1:
Self-driving cars; in 2015, Elon Musk said self-driving cars that could drive “anywhere” would be here
within two or three years. did not happen yet.
Example 2:
Artificial intelligence; in 2015, Mark Zuckerberg said “one of Facebook’s goals for the next five to 10
years is to basically get better than human level at all of the primary human sense: vision, hearing,
language, general cognition.” also not very realistic, a bit over-the-top.
Panic: when people are overly worried about some new technology.
Example 1:
Bicycles: in 1894 the New York Times proclaimed: “there is not the slightest doubt that bicycle riding,
if persisted in, leads to weakness of mind, general lunacy, and homicidal mania.” also did not
happen.
Example 2:
Filter bubbles: in 2011, Eli Pariser introduced the concept of filter bubbles: people becoming isolated
from other opinions, because news & social media algorithms feed them only items that they like and
agree with. Is this a real issue? Or another unwarranted panic? Need to research that.
Hypes and panics might form a feedback loop:
Questions for this course:
- How can we find the right balance between hype and panic?
- How can we assess the true effects and capabilities of smart technology?
Historical context
Short history about AI:
1950: Alan Turing publishes his “Computing Machinery and Intelligence”, proposing what is
know as The Turing Test.
1956: The Dartmouth Summer Workshop on Artificial Intelligence.
1966: Joseph Weizenbaum publishes ELIZA, the first chatbot.
Until 1970s: optimism!
1973: Lighthill report (skeptic about AI), start of first AI winter (period in which people where
negative about the idea of AI, lasted about 8-10 years).
Early 1980s: careful optimism again.
Late 1980s, early 1990s: second AI winter.
, 1990s, 2000s: rise of Machine Learning.
2010s: rise of Deep Learning.
Mid-2010s – Now: optimism … but until when?
Technology is never neutral
Technologies are never neutral. They can for example benefit one group in the society, but another
group of people in the society is not able to use it.
Example 1: Hearing aids
Pro: better hearing, connecting with others.
Con: hearing unwanted sounds, blamed for disability.
Without technology: patient.
With technology: ‘unwilling to do anything about it’.
Example 2: Soap dispenser (with sensor)
Pro: Improved hygiene.
Con: Darker skin colors are not recognized by some of these sensors.
So, the way that technology is designed impacts who gets to use it. Therefore, very important to test
technologies on multiple kinds of people, situations, etc.
Example 3: Image recognition systems
Pro:
Improved searchability;
Automation of tedious tasks (e.g., going through
all your pictures).
Con:
Could lead to an invasion of privacy;
Labels impose/encode a particular world view:
which objects are relevant?
o influences follow-up work: people
often study/use what is easy to
study/use (e.g., focus on Twitter data,
because it is easy to get).
Examples of language technology
1. Chatbots;
2. Voice assistants;
3. Grammar check, autocorrect;
4. Search engines/recommender systems (e.g., Google, YouTube, Spotify);
5. Sentiment analysis;
6. Spam filtering, content moderation.
Questions to ask
- What happens if it works well?
- What happens if it doesn’t?
- Who is this system likely to be successful for?
