Friday, April 28, 2017

unfairness in automated decision making in society.

reading this book about mis-use of maths/stats recently, i think we can go further in condemning the inappropriate approach taken in some justice systems to decide whether a guilty person receives a custodial sentence or not.

The purpose of locking someone up (and other stronger sentences) is complex - it can be to act as a disincentive to others; it can be to protect the public from that person re-offending; it could be a form of societal revenge; and it might (rarely) be an opportunity to re-habilitate the offender.

So we have a Bayesian belief system in action, and we have a feedback loop.  But we better be really careful about i) the sample of inputs to the system and ii) the sample of outputs....and not forget these are humans, and capable of relatively complex and highly adaptive behaviours.

So what could be wrong with the input? (sigh, where to start) -
people who commit crimes are drawn from a subset of society, but people who are caught are drawn from a biased subset - firstly, they're probably less well educated, or dumber, or both, because they get caught. secondly, they're probably from a socially disadvantaged group (racial minority).
people who are found guilty are also the subject of selection bias (and people who get away with it, are party to survivor bias too) - juries have re-enforced the bias in the chance they are caught.

people who are sentenced acquire new criminal skills - this may make them less likely to get caught if they are just poor, but more likely if they are dumb.

So in there' I count at least 4 ways that a decision system that looked at re-offending rates, and properties of the person found guilty, would be building in positive feedback that will lead to more and more people being incarcerated, with less and less justification.

occasionally, external changes (accidental natural experiments) perturb the system and make this more obvious - in the film documentary, the House I live in , the absurd war on drugs is shown to be massively counter-effective - near the end, the huge bias that this has set against african americans starts to wane, simply because of the move in the poor white working class of america into making and consumption of crystal meth (so brilliantly portrayed in Breaking Bad - suddenly, the odds stacked against on group, multiplied by re-enforced prejudice 3 or 4 times over (indeed, one more time for the 3 strikes rule), hit lots of "trailer trash"....

An interesting research task would be to run a model inference tool on the data and see how many latent causes of bias we can find - maybe my 3,4 or 5 is not enough.

truly the world is broken, when it comes to evidence based decision making!

Saturday, March 25, 2017

of the internet, for the internet, by the internet

what have we wrought?

i don't think it is about the echo chamber, bubble, or
faddish claims about fake news and alternative facts.

nor do i accept that  the internet offers a zero-cost channel - the internet switched the value-propositions around by reducing cost for sender, but for some kind of content, it simply moves the cost somewhere else

1/  to the receiver (spam/advert/recommend, whatever you call them) -
2/ to the content creator (for music, film.games etc)
3/ to regulator (to ensure neutrality, control monopolistic tendencies etc)
4/ to the service provider as real competition drives profits to truly marginal
5/ somewhere we havn't thought of yet

so what we didn't think about was how to design robust games to allow people to design and choose appropriate system architectures for sustainable worlds, whether journalism (that doesn't let the vocal extreme minority control the agenda) or creative industries (so original work is rewarded), or peer-economic structures like uber, airbnb, etc that treat the means of production/labour force fairly...

hard times

[yes, i know this is sort of a version of jaron lanier's stuff, but it is becoming more and more evident that the complaint is right, but we need an actual fix, and that that is the hard problem, not identifying the cause, but designing the solution]

Monday, June 13, 2016

Five Digital Epistemological Objects for 2064

Five Digital Epistemological Objects for 2064
or
A Digital Narrative Ark of the Knower

jon crowcroft, Cambridge, 8.5.2014


dreams, visions and prophecies in bits

It is too hard for humans to fully comprehend humans, but it may be possible
to construct a digital model, a computer simulation or even emulation,
that is accurate, not just descriptive, but also predictive. Such a
model would embody modes of thinking that are not entirely rational,
which is what current "AIs" attempt, but would extend to domains
which, I believe, are entirely human, such as dreaming and visionary
or prophetic processes - these are not magic, or pseudo-science ideas,
but ways in which human thought processes leapfrog piecewise or
incremental steps, perhaps building on such mundane stages, but only
revealing themselves thus-wise, as revelations. Not blue gene beating
humans at chess, but more surprising.


computational ethics

We struggle with ethical dilemmas. Why? there are ambiguities or
paradoxes. These are quite easy to express in the right formal
systems, so we should be able to create, perhaps with help from
machines, ethical props, crutches, to help guide us to what is right.
Asimov laws of robotics (4 in the end) were naive, but a start - we
should play with more such. The history of robots (golems, rossum's
universal, mary shelley's etc) is littered with great examples.


diseases who think

it is a high pomp of pretentiousness that only humans think. we know
(e.g. from Dunbar's (and Alison Richards') studies of apes)
that the theory of mind is present to some degree in other creatures,
and sometime, less so in some people.
But the most alien of creatures, such as hive animals, and,
in extremis, bacteria are capable of collective reasoning. Can we
train them to help us? Can we
infect people with thoughts, literally, rather than merely
figuratively?

Bring meaning to Pat Cadigan's notion  of being incurably informed
(see Synners).


haunts - memories stronger than reality

smells, and superstitions, influence us and resonate more than careful
abstract recollections. Perhaps there's an embodiment of knowledge in
these modalities that we could build better, artificially, than
already exist. Can we code ghosts?

learning to un-banish ghosts might be the ultimate rationalisation.

embodiment of knowing in the knower, is in some cases physiological
(scent, muscle memory, belief) - capturing this missing element (where our
typical current digital media representations address typically only 2
or 3 (sight, hearing, perhaps touch) of the more boring senses, seems
like a worthy goal in terms of understanding our understanding more
deeply.


frailty -

we need digital analogues for flakiness  - just as digital
transmission of moving pictures can "degrade gracefully", perhaps
knowledge can be coded in ways that can still be usefull when partly
rotten - as with the human suffering from dementia, still able to carry
out some cognitive tasks, perhaps artificial thinking can be made
resilient. [today's programs, if even slightly corrupt, simply work
then fail - this is a poor show].

In a deeper sense, reflection on the inherent inaccuracy of representation
is needed, etc

indeed, the optical metaphor can be (over-)extended, using the notion
of different lenses, not just for different viewpoints (different
epistemic architectures) but also for level-of-detail - zooming in to
some (reductionist) model, or retreating to some level of abstraction.
Technology (that is processable - i.e. usually digital) can help with
this - indeed, statistics, visualisation, modeling in general, or
towers of models, are all about this.

losing detail is not necessarily loss of knowledge - indeed, the
ability to ignore detail (see the wood for the trees, or the aforesaid
abstraction process) is one of the more useful human (cognitive?)
skills.


--------> Notes and Websites

The mantra data -> information -> knowledge -> wisdom
(c.f. tofler and brunner's future shock/shockwave rider)
is glib, but useful. each stage in this notional process adds
some sort of structure and processing, whose algorithms and
representational choices are themselves just more data (as per the
Eckert/Von Neumann Stored Programme Computer Architecture - sometimes
incorrectly ascribed to Alan Turing:)

Provocations from the meeting of 7.5.14 at CRASSH:

Q.what diff does move from analog to digital make w.r.t knowledge?
[not restricted to humanities part of digital (humanities)

-ve A
n.a. no change
n.b. networking/
n.c. distributed knowledge

n.d. just scale/efficiency...
[me, but emergence - see below and

http://rappers.mdaniels.com.s3-website-us-east-1.amazonaws.com/
http://www.theliteraryplatform.com/
http://fufufo.com/

two types of DH
1. boring: use of computational tools to do studies like word count in jane austin
2. more interesting - humanities study of social/digital/new media

what about both? e.g. study of sampling/mash up?
or
http://www.digitalgovernmentreview.org.uk

+ve A - changes knowledge & also modes&modalities of knowing...

so not H applied to D, but D  to H
so how do humans change when they go digital...

e.g. measure of time - exact? v. inexact
so exactitude is itself a new suitable topic....

---->
every decoding is an encoding...maurice zapp, in lodge's small world:)

e.g.  science - robot scientists discovery/sharing:)
eScience program (e.g. climateprediction.com, seti@home etc)

better e.g. Maths:
proof assistants
Coq & Isabel
e.g. 4 color map & Fermat's last theorem)

Lessig: code as law

2. quantitative: cost copy -> zero (recall)
[all email since 1976]

Piketty's Capital in 21st Century - 20 countries for 150 years...
Scale sometimes is a qualitative change - emergence

3. qualitative: artefacts...

Bad - ideas (e.g. big data) broken (lose nuance)
good - new forms (susan collins @ slade - many turner prizes...


---->
Discussion:-

+ culture
+ society

Piketty: capital in 21st century - twaddle v. girlfriends...

narrative v. sci method -
just different points in process in science v. humanity work...?

versus! creative step in science is still not understood:)

The mistakes are ... interesting..-slade art...
the two brians (may&cox:)

read also: more than human (theodore sturgeon) and
shockwave rider (john brunner)

see also post modern object truth & no value judgements ?:-)
liberal  arts students in 70s who went into west coast startups
may have become unethical coz of this:)

diy:and failure machines:
http://www.katjungnickel.com/2014/02/28/dagstuhls-diy-networking-seminar-making-a-failure-machine/

Saturday, May 28, 2016

cats will 0wn the Internet of Things just like the rest of the Internet

we have a smart cat flap - from a jolly good company called sureflap. we've had it a few years. Our cat is chipped so if she gets lost she can be returned and so the vet can tell what treatments she's had etc etc - all good...

we live in a crowded cat neighbourhood, so many cats try and come into our house to eat our cats food and generally invade her space etc

so we got this cat flap as it reads pet chips, and can be programmed for a given one (actually, a bit like your WiFi AP, for which see more later, it can store up to 30 cats RF-IDs - jolly good, so far).

So then the cat flap goes wrong (starts running batteries flat every day-  normally they last nearly a year....so we go on the company website, and they have a neat diagnostic tool, and we run through it and they say it needs replacing (the smart flap, not the cat:-), and we enter the serial number (of the flap, not the cat) and they say "yes, that is still under warranty and they are sending us a new replacement (very smooth service indeed- arrives next day!).

so we install replacement asap as I am getting fed up with old one using up so many batteries, but I am in a hurry to get to work, so I put the cat flap in the default learning mode, which is that it flashes its little light once a second until the first cat goes through, at which point it stops learning and only lets that particular cat-id in/out.

so i get to work and there's a frantic phone call, and someone tells me some other cat has come into the house first, and eaten our cat's food, and now, only it can get in & out and our cat is stuck in. oops.

so you have to ask how did the alien cat know to try  just then? I mean we know which cat it is and its lived around here for 5 years and it must know it couldn't get in thruogh the old catflap, so what told it that there was a new one? cunning eh.

two things - 1 there is a different learning mode which only leaves a 10 second window, but you have to have a tractable cat that will oblige and train the flap on demand - hard to do. there isn't a way to "migrate" the old cat learned IDs from an old flap to the new one (the way you migrate your contacts lists from old phones to new ones) which would be neat, especially if you had 30 cats! waiting to train all of them could be like, errrrrr, herding cats :-)

on the other hand, alien cats will have it purrrr0wned in 9 1/2 seconds!

Friday, May 27, 2016

credentials, careers and punctuated equilibrium

life is like a sequence of flights where there's an exciting (and unnerving) takeoff (often preceded by stressful and boring waits) followed by the moment you break through the clouds, and the plane levels off in the light, and coasts. the metaphor seems to fit school, college, job changes, partners, kids, deaths/bereavements in family, etc

so purely from a work perspective, this applies to some research projects i've done.....

most the 1980s, we were building/measuring/optimising the basic internet (both on paper, similation, and real code and networks) - culminated in multicast, tcp congestion control, satellite access (in 1988), which smarter people at the other end of the net wanted to test, so we were happy to be on this end of those tests....

then in the 1990s we were doing multicast - both applications (games, vr, and most interesting, Reuters realtime share trading network), and realtime multimedia (Internet TV - what became the main way AT&T, Telefonica and Virgin/NTL built their TV streaming service; and internet telephony/conferencing-  with video, audio, shared whiteboards - etc - what became skype, webex, etc).

then in the 2000s, we were doing opportunistic networking (community mesh, also) + cloud + social media analytics....how well does a kickstarter campaign work? how do people find or follow unbiased news on twitter etc.

now what? I guess its either data science (inferencing latent variables and models) or internet of things, or both, or neither....

one underpinning theme is decentralization - the early internet was, and cloud was meant to be - so now we're revisiting both the wireless net and the cloud to see if we can make them work better without centralization and loss of privacy. See this talk for why&how

Oh yes, why "punctuated equilibrium"? because basically that's another metaphor for what happens in evolution, applied to ideas - change of environment, leads to specation. selection/crossover leads to new ideas and refinement. next....

Wednesday, May 18, 2016

Every Day Data Science Challenges

1 Guitar Strings

Different strings break at different rates. You can buy them singly or in sets of 6 (occasionally with a spare top E) - really what you want is just in time delivery of a new set wtih a distribution of strings (EBGDAE) that matches the wear/tear rate for you, your guitar (classical, flamenco, acoustic, electric etc) and tone/newnewss you like - this could be crowdsources by instrumenting tuning apps on phones which would notice when you tune from way below (e.g more than a 5th below the right note for that string, probably indicates a new string being put on) -

The statistics could be aggregated, and classes of users found, and then companies (like my fave ) could build orders for you -

2 Bicycle Wheel Spokes

I've lost 4 spokes over the last 5 months cycling in Cambridge - probably, they went on the appalling potholes on station road, or the tree roots across burrel's walk - wouldn't it be nice to know where these occcurred so I could report them to the council (and get money:-)

This could easily be done with accelerometers in smart phones....and GPS - look for rapid up/down movement - then afterwards (when a spoke has gone) you should be able to find the periodic wave of the bike as the wheel is now eliptical....

Thursday, April 07, 2016

Miro's Law


To his excellency, High Admiral, 1st Fleet, Third Arm Galaxy 74434
From Captain Moore, First Contact Team, outer quadrant 2,
planet three, yellow star 17.

Your Excellency,

I am writing to report on the results of the strange mission we have undertaken exploring the deserted planet full of wonders known by its former inhabitants, as far as we can tell, as Silt or perhaps Clay - our translators are working flat out to improve on our understanding as there is so much potentially to be gained from their technology and civilization.

We have uncovered a most exciting cache of documents which may finally explain the departure of the beings who constructed these wonderful buildings and devices. Everywhere we look there are suitable perches on tall polls, connected by long nesting strings. Marching across the countryside between what we believe to be their latrine sites, are long flat landing strips where competition for suitability for breeding stock amongst our warriors would be idea. Enough of the marvels. Back to the documents, which appear to be from a savant, desperate to solve some crisis that has struck their ecosystem....he writes

"I am sending you this letter to ask for your advice. As I travel often to far away lands, and present my work at conferences, I am inevitably showered with gifts, and amongst these is always at least one ball point pen. Since we developed the fishnet" (we're not sure if that is the right term) " I have had almost no use for these gadgets, and have steadily been accumulating them in killing jars" (again, not clear, but it seems that the 'jar' is something to preserve things in at least we can tell that). "Now, I am out of jars, and am unable to store sufficient foods for the winter" (now you see why we are fairly sure of our translation).

"Hence I am writing to all my fellow savants, to ask if they have any idea how we can solve this problem. The number of ball points is growing hyper-exponentially, and threatens the whole of the Fellowship of the Royal Society, and so we realise that we have to reach out to our cousins across the seas for help

I remain yrs, etc etc

p.s. I enclose a pen for your use in response"


Your excellency, there are many responses, but most echo the concern, and dare I say despair at the situation. Finally, however, after much work, it seems that one amongst these giants of intellect proposes a possible solution, for it is the last letter, and in it are many strange symbols which resemble our own formalism for hyperspace drive, and yet appear to arrive at a simpler solution....the letter concludes

"and so we suddenly understood that if we could figure out how to channel the gravity waves just right, the spaghettification phenomenon will allow even the largest of our fellow humans to fit down the inter-dimensional tunnel formed from the tubular casing of the ballpoint through the frames, and into the landscapes that miro draw, it seems from true life, rather than, as critics of the day said, his dreams. the rest can easily be worked out from this sketch...

I leave now, as you know, however, I enclose your last pen, which you will find sufficiently charged to allow you to join us in Blue II should you wish, although as you are of hungarian origin, you may resonate with the richer Mikrokosmos..."

So, your excellency,
It seems that the "humen" of "mud" were finally able to lose enough weight through some fantastic new plan, to soar to other dimensions.....we have yet to complete the proof from the sketch mentioned above, but I hope that we can report to you on this soon,

I remain, as ever, your nest-issue of the fourth degree, captain Moore...

captain etc

p.s. I enclose a pristine unused biro pen for your perusal, and who knows, perhaps use....

Blog Archive

About Me

My photo
misery me, there is a floccipaucinihilipilification (*) of chronsynclastic infundibuli in these parts and I must therefore refer you to frank zappa instead, and go home