20191009 - LASCOT2019 - In defence of uncertainty, Abeba Birhane @Abebab

by Thierry de Pauw on

#lascot

In defence of uncertainty, Abeba Birhane @Abebab

Cognitive Science PhD candidate
School of Computer Science, University College Dublin

What I am going to talk about, is what I am trying to write for my PhD thesis.

I'm not totally sure the talk is going to be about uncertainty.

we love certainty! we have always sought certainty.
=> secure foundation for knowledge in the face of radical doubt

Rene Descartes (Franch mathematician and philosopher) "Cogito ergo sum"
has been obsessed by certainty

  • @somesheep: We love order, it gives us control. We strive for certainty in Western Philosophy. Descartes warns us that our sense can deceive, so we need to doubt to assess information.
    @Abebab #lascot https://t.co/Lz5NrYZSEw

  • @somesheep: The one thing Descartes was sure of was that there is something doing the thinking.
    @Abebab #Lascot

"Traditional science in the Age of the Machine, tended to emphasize stability, order, uniformity, and equilibrium. Whereas most of reality, instead of being orderly, stable, and equilibrial, is seething and bubbling with change, disorder, and process" -- Prigogine & Stengers, 1985 @Abebab

  • @cyetain: The sciences always carry within themselves the project, however remote it may be, of an exhaustive ordering of the world —Foucault

lascot https://t.co/HX0Ya81Y86

To order the world, is a political act. There is always a power-relation. @Abebab

  • @cacorriere: RT @somesheep: But reality is not like a machine and striving for certainty is self-delusion and categorising (ordering) the world the way science does is political. See etymology of “science” (dissecting). @Abebab #Lascot https://t.co/JU2Sx2jaqy

One of the most highly developed skills in contemporary Western civilization is dissection: the split-up problems into their smallest possible components. We are so good at it. So good, we often forget to put the pieces back together again. -- Prigogine & Stengers, 1985

Good example: neuroscience
dividing the brains into neurons, the neatly fine details
but what does it mean for that person?

  • @somesheep: Neuroscience has delved into the fine-grained details of the brain, can explain how Neurons fire, but has nothing to say about what this means for us as people. We lack skills of zooming out, seeing the whole and start making sense from there. @Abebab #Lascot

In place of messy, ambiguous, and continually changing reality
=> what we do is: we try to stabilise, work in silos, strive for order and the "view of nowhere"

  • @somesheep: So we cling to organising the world in our minds. We claim to capture order as objective representation, and fail to see that attaining certainty is not only unattainable, it should also be undesirable. @Abebab #Lascot https://t.co/L2yrV1G9ge

and that perspective is seen is being universal.

and we think we are objective

Objective is a subject's delusion that observing can be done without him. Invoking objectivity is abrogating responsibility - hence its popularity. -- Heinz ...

Is absolute certainty attainable? Desirable? I'm trying to convince you "No". @Abebab

  • @somesheep: An alternative view is offered by Ubuntu (the southern African Philosophy, not the OS). In Ubuntu you exist because you are interacting with others. Existence means co-existence. Instead of “I think, therefore I am” it says “a person exists through other persons”.
    @Abebab #Lascot https://t.co/dOhfpggGsm

  • @somesheep: Existing is a response to others. In isolation, we lose our sense of self. We can observe this with people in confinement, who come out having lost their ability to be rational, loss of sense of time and place.
    The “idea of me” is constantly moving and changing.
    @Abebab #Lascot https://t.co/GAyizOvLc7

Multivalent

The self I take grocery shopping, say, differs in the actions and behaviours from the self talking to my PhD supervisor.

  • @somesheep: I will be different depending on who I talk to and where I am, different on stage from when you can chat to me later, different when talking to my PhD supervisor and different again when being with my family. @Abebab #Lascot https://t.co/OQWtpeYnXC

As we strive towards certainty and stability, we quantify and formalize relational, multivalent, ambiguous, unquantifiable, dynamic, continually moving, context and history dependent human attributes

and technology increasingly plays a significant role

Technology is directly related to Western thinking.
=> technology solutions to anything and everything is rampant

  • @sophiedennis: Technology drives us towards certainty and stability (it’s what tech is good at: rationalising the complicated), and so to rationalise what is inherently relational, ambiguous, dynamic and unquantifiable - @Abebab #lascot https://t.co/I6d0mKbcvw

Examples:
Can face classifiers make a reliable inference on criminality?
An app that can select the perfect babysiter using AI

  • @malk_zameth: «we tend to look at technology as the answer to everything» — @Abebab #lascot Abeba showing cases of we trying to use technology to define “what is a good babysitter”,”what is consent”, “who will be a criminal” All unquantifiable things https://t.co/Q0tvIhnERZ

Quantify the unquantifiable.

Build tools that feed into a system of punishment.
=> It is very likely that we are doing more harm than good.

  • @somesheep: We push things which are unquantifiable into simplified and stable apps. When we do that, we are likely to do harm and the tools we use hide that from the creators, but especially when we want to punish, we think that’s acceptable.
    @Abebab #Lascot https://t.co/RRpA38TSQE

  • @thecodecleaner: #lascot when tech learns it can only make decisions based on what its trained on. If that doesn't include women being hired, it won't hire women. @Abebab

=> trying to create a future based on the past => introduces bias

Examples:
- Amazon scraps secret AI recruiting tool that showed bias against women
- Women less likely to be shown ads for high-paid jobs on Goggle, study shows
- Google apologises after app tagged black people as "gorillas"
- Machine Bias: software used to predict future criminals and it's biased against blacks

(see photo for more examples)

  • @kenny_baas: #lascot @Abebab we feed AI with data from the past, building in current status quo. This way white male keep the power and women and coloured people keep being suppressed.

  • @SteveSmithCD: RT @CatSwetel: Wow, @Abebab is helping me realize how some of the Agile practices we advocate can lead to mechanistic thinking. There's such a temptation to break work down into the smallest possible units rather than the smallest useful/valuable units. #lascot

"When we try to pick up anything by itself we find it is attached to everything in the universe." -- John Muir

Weizenbaum (1972): The structure of the typical essay on "The impact of computers on society" is as follows: ... (see photo for more)

There exists no technology without any side-effects.
see @sarahmei tweet thread

All models are wrong; some are dangerous and harmful. @Abebab

"Purely technical solutions" to issues that intersects with various social, cultural political, historical moral cultural ream means we are blind to complexities and very likely to do more harm than good.

  • @somesheep: Technology is in itself bad or harmful, but we need to stop just checking if models are useful (George Box), but make also learn how they can harm, before we deploy them. A consequence is that technology-only solutions for social problems cannot work. @Abebab #Lascot https://t.co/Yz9Ewq3I8c

We must acknowledge that we are part of the world we observe, quantify, measure and model. @Abebab

"I'm just an engieer"
(February 2018 Researchers from UCLA presenting a paper on predictive policing system)
=> this impacts society

  • @thecodecleaner: #lascot an observer affects the system it observes @Abebab https://t.co/L0seHgvLbg
  • @sophiedennis: Is it acceptable to say “I’m just an engineer” (or a designer) when asked about ethical and social implications of our work? - @Abebab #lascot https://t.co/OETKZcjykF
  • @somesheep: If you enter a social realm ignoring everything connected except an engineering task, you are abandoning moral responsibility and you alter the social fabric towards treating people as machines. @Abebab #Lascot https://t.co/JlFAYBsCuh

Any model of the world that we construct has moral and ethical dimension on it. @Abebab

"Systems of classification are themselves systems of power", Stuart Hall @Abebab

  • @CatSwetel: Some technology just shouldn't exist in the first place. - @Abebab #lascot
  • @kenny_baas: @Abebab #lascot putting it together with George box quote "all models are wrong, some are useful, some are dangerous. Because with models you abstract and leave stuff out. It is with everything in life, If you include, you by default exclude. There is no pure positive tech. https://t.co/tG49ax7uur

ASK: "Does a technological intervention result in more trouble or harm than the situation it's meant to address?" -- Baum ... (missed the author) @Abebab

  • @CatSwetel: RT @sophiedennis: .@Abebab ending powerfully “some technologies should not exist”. Ask yourself why does this thing need to exist? Why do we need to know? How might it do harm and is whatever vague upside we think it offers worth that? #lascot

"The price we pay for the potential of true novelty and creativity is uncertainty. Uncertainty in its very form is a negative word. It should not be. A universe in which certainty ..."

Absolute certainty is not only unattainable but also dangerous. Reality is messy. @Abebab

Takeaways:
- It's not about the 'right answers' but asking meaningful questions
- Push more towards understanding and less towards prediction
- In place of final, and fixed answers, we aspire for continual negationation and revision
- In place of certainty, we emphasise indeterminability, partial-openness and unfinalisability
- Partial-openness leaves room continual dialogue, negotiation, reiteration and revision