Dominic Cummings is part of Sage

[pontypridd]

The mystery scientific advisory body

[LilacTree1]

It’s not news - I just thought DC wouldn’t crash the economy but clearly that was stupid of me.

[AnneElliott]

Everyone's entitled to their opinions of course. But some opinions are better informed than others.

Very interested to hear from anyone who's worked at senior levels of the civil service (under any Government ) that thinks this is either an issue or a change from previous Governments.

[Humphriescushion]

Absolutely some peoples opinions more informed still allowed an opiniion though.

[C8H10N4O2]

How does the use of this data affect us on an individual level?

I didn't get back to the thread yesterday but Perking covered much of what I would say. At an individual level the exposure of private information in government datasets such as medical records is significant. There are special exemptions granted from data privacy laws around using much of this data. Ben Goldacre was right to be so concerned.

Even where data is "anonymised' and sold quite openly the data is still identifiable. Commonest form of anonymisation is to remove name and street address but leave the postcode. A UK postcode has a maximum of 50 addresses, usually less. How difficult would you find it to identify the 45 yr old Sikh BAME man living with a female partner, a child under 10 and a mortgage from 50 addresses? Then merge that data with SM feeds, marketing and customer feeds etc. Then add in telco data from the trackign devices we all carry.

Ten years ago we could predict pretty accurately what any one person would be doing at 10:00 the next day. What we couldn't do was process the data at scale across the population.

The outputs enable very targeted micro advertising/messaging. So for a single message you can tailor a hundred versions each of which might only go to a few hundred people. You can microtarget individuals to influence their decisions on key committees and even juries (juror's names are not secret). Its very difficult to identify or prove and its extensively used both in marketing and populist politics where there is enough money behind it.

[C8H10N4O2]

Government advisor attends meetings - shock... Er, that is what they do. Stop looking for things that aren't there.

Its not the norm for SAGE, its especially not the norm for an unelected political advisor with an established agenda when in a critical situation. I'd be interested to know what the level of meeting was when Sorene claims to have attended on behalf of the chief apologist for Facebook. (ie was it a SAGE briefing meeting to the government rather than the core SAGE meetings).

SAGE isn't a cabinet select committee - the whole point is it is supposed to function independently of the politics. The SPADs then advise on the politcal implications of the advice.

[C8H10N4O2]

If the list of members in SAGE and those observing is published, then there is transparency and it's not a an issue.

Quite.

[C8H10N4O2]

Steve Hilton, Cameron’s highest-profile adviser from that time, left government in frustration after two years.

That would be Steve Hilton who actually left for a lucrative career on the West Coast and currently has his own Fox News channel advocating the Trump line of open up the economy and the consequential deaths are simply collateral damage.

[C8H10N4O2]

Another Dominic Cummings' plot?

SOrry I've no idea what point you are trying to make here. There is nothing in your links which is inconsistent with my statement.

[PerkingFaintly]

You can microtarget individuals to influence their decisions on key committees and even juries

Yes! The Good Fight did an episode on this!

4 min excerpt of Marissa being clever. Which is always fun to watch:

[C8H10N4O2]

Yes! The Good Fight did an episode on this!

Yes I remember that - I had a bunch of people asking me if it was science fiction .

I was surprised that it had made it into mainstream drama, the depth of the targeting is not that well known outside of the industry and a subset of its customers.

[C8H10N4O2]

Of course the other problem with this large scale data collation and automated decision making is that it also reinforces existing prejudice and assumptions.

For me that is one of the greater concerns at the moment. "Computer sez no" is bad enough, when the people saying it actually believe the computer knows better because "machine learning" its actually quit scary, especially in medicine.

[PerkingFaintly]

Indeed, C8H10N4O2. (We clearly do move in the same information circles...)

Good explanation of some issues with machine learning:

This is how AI bias really happens—and why it’s so hard to fix
www.technologyreview.com/2019/02/04/137602/this-is-how-ai-bias-really-happensand-why-its-so-hard-to-fix/

Case study of specific piece of US software:

Machine Bias
There’s software used across the country to predict future criminals. And it’s biased against blacks.
www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing

[Longtalljosie]

He’s going he government’s chief communicator. It’s a committee about how to get the general public to behave in the optimal way in a pandemic. Of course he’s bloody on it.

It’s clickbait journalism of the worst order. Because anyone in journalism with two brain cells to run together would have bet the mortgage he had to be on it. So confirm it then start pretending to pearl-clutch. Disgraceful.

[PerkingFaintly]

I agree that human behaviour faced with a machine decision is even more worrying.

It can be very hard for a person to act in contradiction to what "the computer sez".

In the first instance, a lot of people will just want to believe the computer – it's given them the answer, why think any further?

And in the second instance, a person who is pretty sure the computer has given the wrong answer is nonetheless faced with a dilemma. If they follow their own judgement, and everything doesn't turn out absolutely perfectly, they will certainly be held culpable and punished much more harshly than if there had been no computer involved (even if some authority acknowledges the computer decision was very likely wrong).

So in medicine, or policing... well, how could that possibly go wrong...?