Share this page!

Rachel Thomas Profile picture
Twitter logo
29 Nov, 10 tweets, 6 min read
My newest post: Doing Data Science for Social Good, Responsibly 1/

fast.ai/2021/11/23/dat…
Data use by non-profits can be powerful, such as this CA vaccine equity analysis by @ACLU_NorCal @snowjake & @DataInstituteSF @MaxShinnerl (I mentored) 2/

Recent projects in which @QUTDataScience partnered with @CCQld & @fareshare_aus were a success 3/
However, there are pitfalls to avoid when approaching "data for good." As @sarahookr has pointed out, it is often an imprecise catchphrase 4/

towardsdatascience.com/why-data-for-g…

Some "data for good" projects (such as tracking people through cell phone data) raise issues of:
- lack of consent
- risk of privacy breaches
- potential for misuse
- are results even being used?
- are their less invasive approaches? 5/

nature.com/articles/d4158… @amymaxmen
It is crucial to recognize power differences among various stakeholders 6/

"Social good projects should be developed at a small scale for local contexts — they should be designed in consultation w/ the community impacted by the systems in order to identify core values & needs.” 7/

medium.com/a-new-ai-lexic… @AINowInstitute
Advice on data for good:
- be specific about goals & impact on different stakeholders
- center the problem, not a flashy solution
- take privacy seriously
- include people most impacted & recognize their values may be different from yours
- avoid answering the “wrong problem” 8/
Also, consider practical processes you can implement to make ethical reflection, deliberation, & judgement more concrete in data, engineering, & design workflows, such as @scuethics @ShannonVallor Tech Ethics Toolkit 9/

scu.edu/ethics-in-tech…
More in my full post here 10/
fast.ai/2021/11/23/dat…

• • •

Missing some Tweet in this thread? You can try to force a refresh
　

Keep Current with Rachel Thomas

Rachel Thomas Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @math_rachel

Rachel Thomas Profile picture
28 Nov
Flaws of countering disinfo w/ appeal to authority:

"Worrying about whether we trust institutions without asking if these institutions deserve trust... A program of infantalization – trust that the adults know what is right – will provoke equally infantile resistance." @Aelkus We have gotten very far from the original goal of trying to
Failure of legacy institutions to respond appropriately to the pandemic, from March 2020 @aelkus, h/t @RSButner

aelkus.github.io/problem/2020/0…
A society that cares more about declining trust in institutions than what institutions have done to deserve trust – and which devotes far more effort towards managing the behavioral psychology of risk than actually reducing risk – is engaged in narrative-making above all else. Managing public health and disease was one of the core tasks
Read 4 tweets
Rachel Thomas Profile picture
26 Nov
Compared to ethics principles in medicine, AI ethics principles lack:
1. common aims & fiduciary duties
2. professional history & norms
3. proven methods to translate principles into practice
4. robust legal & professional accountability mechanisms

nature.com/articles/s4225…
"The truly difficult part of ethics—actually translating theories, concepts & values into good practices AI practitioners can adopt—is kicked down the road like the proverbial can." @b_mittelstadt 2/
"Ethics has a cost. AI is often developed behind closed doors without public representation... It cannot be assumed that value-conscious frameworks will be meaningfully implemented in commercial processes that value efficiency, speed and profit." 3/
Read 4 tweets
Rachel Thomas Profile picture
25 Nov
🧵on Google, Tech Regulation, & Risk of Industry Capture in 🇦🇺

"Google’s investment in Australia’s tech future — a boon for universities or a Faustian bargain?" by @jordanguiao 1/

abc.net.au/religion/googl…
Australia's competition regulator found:
- Google engages in anti-competitive behavior in digital advertising, which harms consumers & businesses
accc.gov.au/media-release/…

- monopoly on search has "stifled innovation and reduced consumer choice"
ia.acs.org.au/article/2021/a… 2/
Australian federal court found that Google misled Australian mobile and tablet users on how it collects location data 3/

abc.net.au/news/2021-04-1…
Read 7 tweets
Rachel Thomas Profile picture
22 Nov
Many people have a false dichotomy that you are either FOR or AGAINST covid restrictions, with no nuance about the TYPE of restrictions or level of effectiveness, much less that eschewing all restrictions → hospitals collapse & lockdown more likely. 1/
There has been a lot of terrible public health messaging & contradictory government policies in the West, from the start of the pandemic, continuing now, and these erode public trust, create false expectations, & contribute to “pandemic fatigue” 2/
The “only elderly & chronically ill are at risk” was both false AND ineffective messaging. This has been clear from the VERY START of the pandemic. (I RTed @jenbrea at the time) 3/
Read 16 tweets
Rachel Thomas Profile picture
16 Nov
The false hope of current approaches to explainable AI in health care: current explainability approaches can produce broad descriptions of how an AI system works in general, but for individual decisions, the explanations are unreliable or superficial 1/
thelancet.com/journals/landi… The false hope of current approaches to explainable artifici
Explainability methods of complex AI systems can provide some insight into the decision making process on a global level. However, on an individual level, the explanations we can produce are often confusing or even misleading. @MarzyehGhassemi @DrLaurenOR @AndrewLBeam 2/ What are explanations for? These limitations do not render e
Increased transparency can hamper users’ ability to detect sizable model errors and correct for them, "seemingly due to information overload." 3/ The intuitive simplicity of inherently explainable models is
Read 6 tweets
Rachel Thomas Profile picture
13 Nov
"Who benefits from data sharing in Africa? What barriers exist in the data sharing ecosystem, and for whom? If much of the data sharing practice is shaped by the Global North, how can we ensure that the narrative for Africa is controlled by Africans?" 1/

arxiv.org/abs/2103.01168 Narratives and Counternarratives on Data Sharing in Africa R
Stakeholders in the African data sharing ecosystem. Those at the top of the iceberg hold significant power & leverage in guiding data sharing practices & policy compared to those in the hidden part of the iceberg. More powerful stakeholders wield disproportionate power. 2/ picture of iceberg. From top: (above water) government bodie
Dominant narratives around data sharing in Africa often focus on lack, insufficiency, deficit.

This framing minimizes the strength, agency, and scientific & cultural contributions of communities within the continent, and overlooks community norms, values, & traditions. 3/ a lack of knowledge about the value of data and training, as
Read 7 tweets

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Thank you for your support!

Follow Us on Twitter!

Like this author's thread?

Get emails when @math_rachel has new unrolls!

Check out other threads from @math_rachel!

Read all threads by @math_rachel

:(