Agents & Artefacts
Deep Learning
Artefacts in the public space
100

Provide an example of a public artifact in a park and explain its political qualities.

A historical figure/the placement of benches, the slope of the stairs

100

Give an example of an AI model based on deep learning.

ChatGPT, Midjourney, etc

100

What are the politics of the bridges leading from New York to the beaches of Long Island

Designed low to prevents poor people (who use public transport) from going to the beaches.

200

What is 'secondary agency' in the context of technology and politics?

'Secondary agency' refers to the capacity of technological artifacts to exert influence or agency beyond their initial intended purpose.

200

What is one consequence of deep learning algorithms trained on biased data, as discussed by Lazovich?

One consequence of deep learning algorithms trained on biased data is that they can perpetuate and even amplify societal biases, potentially leading to unfair outcomes.

200

In what ways do public buildings in urban planning reflect political decisions?

Placement, size, architecture, function

300

How does 'secondary agency' relate to the concept of a 'technological fix'?

'Secondary agency' relates to a 'technological fix' by highlighting that technological solutions to societal problems can have unintended political consequences.

300

Give two examples of high-profile cases of deep learning misuse mentioned by Lazovich.

Discern sexual orientation from facial images, reconstructing a person's face from a recording of their voice, inferring gender from a facial image

300
Explain how the design of a public transport system has political implications.

Placement of station/frequency of transport/ destinations per region.


400

Explain the relation between agents and artefacts.

Agents create artifacts to enhance their abilities and achieve specific objectives. These artifacts, once created, can also influence the behavior and capabilities of the agents themselves.

400

How does Lazovich propose addressing the issue of ignoring the context and meaning of data in deep learning?

Be open about models and add disclaimers explaining where the data came from and what its inherent biases may be.
400

How can automatic soap dispensers be racist and thus have unintended consequences?

Automatic soap dispensers are assumed to be less able to recognise black skin and therefore do not always dispense soap to them.

500

What does the Actor Network Theory (ANT) state?

It rejects a strict distinction between primary agency in humans and secondary agency in artifacts. ANT suggests that both humans and artifacts can be considered "actants" that influence each other in networks of relationships. It emphasizes that agency is not solely human and that artifacts can have intentions or actions within these networks.

500

How does deep learning's reliance on computational resources relate to its political dimensions?

Deep learning's reliance on computational resources centralizes power in the hands of institutions that can afford these resources, potentially leading to the imposition of values and priorities by those institutions.

500

Give three examples on how to prevent unintended consequences of artefacts.

Diverse Stakeholder Involvement/Ethical Impact Assessment/Transparency and Accountability/Continuous Evaluation/Community Engagement/Bias Mitigation/Political Impact Assessment/Education and Training



M
e
n
u