MON: Data Science Jam by Gracen Brilmyer

Patient médical

2024-12-02 14:48:33 UTC

can be biased in terms of how it is collected

2024-12-02 14:48:41 UTC

Can be used for harmful purposes ie predictive policing, employment discrimination etc

2024-12-02 14:49:13 UTC

2024-12-02 14:49:19 UTC

Can perpetuate biases or reinforce them

Observer bias can create sample bias

2024-12-02 14:49:27 UTC

2024-12-02 14:49:35 UTC

entire groups/sections of data may be excluded and therefore the sample that is taken is inaccurate

human are inherently unpredictable/ engage in unpredictable behavior

2024-12-02 14:49:37 UTC

extra data points from biased surveillance can reinforce bias with 'evidence'

2024-12-02 14:49:43 UTC

sample bias

2024-12-02 14:49:50 UTC

Minority Report has become a documentary

2024-12-02 14:49:55 UTC

Predictions can be biased based on issues with data collection and analysis

2024-12-02 14:50:03 UTC

2024-12-02 14:50:10 UTC

implicit biases inherent to whoever is conducting the research/analysis

lack of inclusion/ diversity on back end can cause biases in data

2024-12-02 14:50:18 UTC

Analysis Without Informed User Consent

2024-12-02 14:50:28 UTC

Data can be used for purposes besides the stated intent for its collection, in ways that users may not have been aware of when consenting to share information.

DS passivity scale. Unsure how it works, but interesting to move away from idea that DS is free of bias because it is based on math/technology/stat

2024-12-02 14:50:57 UTC

census data has been used in the past to facilitate genocide

2024-12-02 14:51:01 UTC

can re-identify anonymous people

2024-12-02 14:51:05 UTC

consent is usually not ongoing as pointed out in data manifest-no

2024-12-02 14:51:33 UTC

Current Legal protections

2024-12-02 14:51:54 UTC

Laws like the ADA not presently having any clauses that protect against predictive discrimination or requiring that employers disclose when they are using these models

learned bias influences how models are trained, therefore how those models predict

2024-12-02 14:52:02 UTC

predictive data being used can lead to cyclical data being produced that confirms biases like with predictive policing

2024-12-02 14:52:39 UTC

in medical situations, predictions are unclear when risk factors cannot be easily categorized as genetic or acquired by choice

2024-12-02 14:52:43 UTC

it's like "Surveillance Capitalism"

2024-12-02 14:53:26 UTC

Zuboff writes: if “industrial capitalism depended upon the exploitation and control of nature,” then surveillance capitalism “depends instead upon the exploitation and control of human nature.” (the use personal data subjects individuals to categorization and highlights how people can become voiceless in the face of data misusage)

Presentation of challenges in the last video bring up findings that can be applied in an organization's code

2024-12-02 14:53:28 UTC

people have to pay with their data in order to use ubiquitous technology therefore their consent becomes coerced since there is no alternative option

2024-12-02 14:53:50 UTC

it's impossible to actually read the full terms and conditions for every single technology or website we use time-wise

2024-12-02 14:54:43 UTC

Toward Gataca

2024-12-02 14:55:36 UTC

The way that information is being processed and used to making predictions about the human condition and eventually making decisions discriminate against the inherent differences that exist in the human population. It is a new kind of eugenics?

In the example of overpolicing of certain neighbourhoods, police overcollect data and then use that to reinforce beliefs about the residents' allegedly higher level of criminality

2024-12-02 14:56:26 UTC

this section is v "life imitates art" but only for dystopian scifi:(

2024-12-02 14:57:58 UTC