From cbd057cbf9098dc89f45712591716943d7176bb7 Mon Sep 17 00:00:00 2001 From: Neil Lawrence Date: Thu, 19 Jan 2023 10:16:19 +0000 Subject: [PATCH 01/10] Update for R255 --- _r255/_config.yml | 22 +++++++++++++++++++ ...irness-stochasticity-in-decision-making.md | 1 + 2 files changed, 23 insertions(+) create mode 100755 _r255/_config.yml create mode 100644 _r255/bias-variance-and-fairness-stochasticity-in-decision-making.md diff --git a/_r255/_config.yml b/_r255/_config.yml new file mode 100755 index 0000000..ae89222 --- /dev/null +++ b/_r255/_config.yml @@ -0,0 +1,22 @@ +author: +- given: Neil D. + family: Lawrence + institution: University of Cambridge + gscholar: r3SJcvoAAAAJ + twitter: lawrennd + orcid: 0000-0001-9258-1030 + url: http://inverseprobability.com +layout: lecture +venue: GN06, William Gates Building +postsdir: ../../../mlatcl/r255/_lectures/ +slidesdir: ../../../mlatcl/r255/slides/ +notesdir: ../../../mlatcl/r255/_notes/ +notebooksdir: ../../../mlatcl/r255/_notebooks/ +writediagramsdir: . +diagramsdir: ./slides/diagrams/ +transition: None +ghub: +- organization: lawrennd + repository: talks + branch: gh-pages + directory: _r255 diff --git a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md new file mode 100644 index 0000000..c2695ff --- /dev/null +++ b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md @@ -0,0 +1 @@ +--- title: "R255: Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: week: 1 date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md transition: None --- Welcome to the R255 module on Bi Papers: Performative Prediction The NeurIPS Experiment Variance Modelling English Meaning and Culture Anna Wierzbicka Year: 2006 Chaper 5: Being fair Another Key Anglo Value and Its Cultural Underpinnings \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/trolley-push.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) } \thanks \references \ No newline at end of file From 574270152d0a121c1410ea4461c4e9efcfca6ff8 Mon Sep 17 00:00:00 2001 From: Neil Lawrence Date: Fri, 20 Jan 2023 09:42:07 +0000 Subject: [PATCH 02/10] Update --- ...as-variance-and-fairness-stochasticity-in-decision-making.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md index c2695ff..137a88b 100644 --- a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md +++ b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md @@ -1 +1 @@ ---- title: "R255: Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: week: 1 date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md transition: None --- Welcome to the R255 module on Bi Papers: Performative Prediction The NeurIPS Experiment Variance Modelling English Meaning and Culture Anna Wierzbicka Year: 2006 Chaper 5: Being fair Another Key Anglo Value and Its Cultural Underpinnings \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/trolley-push.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) } \thanks \references \ No newline at end of file +--- title: "R255: Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: week: 1 date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md transition: None --- Welcome to the R255 module on Bi Papers: Performative Prediction The NeurIPS Experiment Variance Modelling English Meaning and Culture Anna Wierzbicka Year: 2006 Chaper 5: Being fair Another Key Anglo Value and Its Cultural Underpinnings \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) } \thanks \references \ No newline at end of file From 125ada208c0a0cab4698ec6da980f1018e2603df Mon Sep 17 00:00:00 2001 From: Neil Lawrence Date: Fri, 20 Jan 2023 09:42:57 +0000 Subject: [PATCH 03/10] Update --- ...as-variance-and-fairness-stochasticity-in-decision-making.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md index 137a88b..533ac12 100644 --- a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md +++ b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md @@ -1 +1 @@ ---- title: "R255: Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: week: 1 date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md transition: None --- Welcome to the R255 module on Bi Papers: Performative Prediction The NeurIPS Experiment Variance Modelling English Meaning and Culture Anna Wierzbicka Year: 2006 Chaper 5: Being fair Another Key Anglo Value and Its Cultural Underpinnings \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) } \thanks \references \ No newline at end of file +--- title: "R255: Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: week: 1 date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md transition: None --- \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) } \thanks \references \ No newline at end of file From 718a9e6a020b43983c30ffdadba1c197912f121e Mon Sep 17 00:00:00 2001 From: Neil Lawrence Date: Sat, 21 Jan 2023 19:21:13 +0000 Subject: [PATCH 04/10] Update r255 with video and more notes --- ...as-variance-and-fairness-stochasticity-in-decision-making.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md index 533ac12..43ca0bb 100644 --- a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md +++ b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md @@ -1 +1 @@ ---- title: "R255: Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: week: 1 date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md transition: None --- \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) } \thanks \references \ No newline at end of file +--- title: "R255: Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: week: 1 date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" youtube: https://youtu.be/lLvzTJtYeU8 blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md transition: None --- \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) } \thanks \references \ No newline at end of file From 9ebd6a2123fba4274d7a7ec2e29c073f2ea4f95a Mon Sep 17 00:00:00 2001 From: Neil Lawrence Date: Sun, 22 Jan 2023 11:27:08 +0000 Subject: [PATCH 05/10] Update r255 with video and more notes --- ...as-variance-and-fairness-stochasticity-in-decision-making.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md index 43ca0bb..7573eb1 100644 --- a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md +++ b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md @@ -1 +1 @@ ---- title: "R255: Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: week: 1 date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" youtube: https://youtu.be/lLvzTJtYeU8 blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md transition: None --- \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) } \thanks \references \ No newline at end of file +--- title: "R255: Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: week: 1 date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" youtube: lLvzTJtYeU8 blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md transition: None --- \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. } \thanks \references \ No newline at end of file From 4ff06f8b5ed8e09a608cc588ad19d5c9c95b60a4 Mon Sep 17 00:00:00 2001 From: Neil Lawrence Date: Mon, 23 Jan 2023 08:06:42 +0000 Subject: [PATCH 06/10] Update r255 with video and more notes --- ...as-variance-and-fairness-stochasticity-in-decision-making.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md index 7573eb1..5bccf9b 100644 --- a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md +++ b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md @@ -1 +1 @@ ---- title: "R255: Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: week: 1 date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" youtube: lLvzTJtYeU8 blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md transition: None --- \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. } \thanks \references \ No newline at end of file +--- title: "R255: Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: week: 1 date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" topic: 7 year: 2023 youtube: lLvzTJtYeU8 blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md transition: None --- \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. } \thanks \references \ No newline at end of file From 1f3d4f07b016f9388d7e106e0ea06de24e1ae95b Mon Sep 17 00:00:00 2001 From: Neil Lawrence Date: Mon, 23 Jan 2023 08:44:14 +0000 Subject: [PATCH 07/10] Update r255 with video and more notes --- _r255/_config.yml | 2 +- ...as-variance-and-fairness-stochasticity-in-decision-making.md | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/_r255/_config.yml b/_r255/_config.yml index ae89222..cc0dee1 100755 --- a/_r255/_config.yml +++ b/_r255/_config.yml @@ -8,7 +8,7 @@ author: url: http://inverseprobability.com layout: lecture venue: GN06, William Gates Building -postsdir: ../../../mlatcl/r255/_lectures/ +postsdir: ../../../mlatcl/r255/_topics/ slidesdir: ../../../mlatcl/r255/slides/ notesdir: ../../../mlatcl/r255/_notes/ notebooksdir: ../../../mlatcl/r255/_notebooks/ diff --git a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md index 5bccf9b..11fbbf5 100644 --- a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md +++ b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md @@ -1 +1 @@ ---- title: "R255: Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: week: 1 date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" topic: 7 year: 2023 youtube: lLvzTJtYeU8 blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md transition: None --- \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. } \thanks \references \ No newline at end of file +--- title: "Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" topic: 7 year: 2023 youtube: lLvzTJtYeU8 blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md featured_image: slides/diagrams/ml/bias-variance008.png transition: None --- \notes{\subsection{Overview}} \notes{In this topic we will explore the relationship between the bias-variance dilemma and real world decisions. The opening session will set the context, then we will read the following four papers.} \notes{1. [Neural Networks and the Bias/Variance Dilemma](https://www.dam.brown.edu/people/documents/bias-variance.pdf) by Stuart Geman, Elie Bienenstock, and René Doursat (the technical material of importance is in Section 3). 2. [Probabilistic electrol methods, representative proability and Maximum Entropy](https://www.votingmatters.org.uk/ISSUE26/I26P3.pdf) by Roger Sewell, David MacKay and Ian McLean 3. [Habitual Ethics? Chapter 7](https://library.oapen.org/bitstream/handle/20.500.12657/58884/9781509920433.pdf) by Sylvie Delacroix 4. [Killing, Letting Die and the Trolley Problem](https://learning.hccs.edu/faculty/david.poston/phil1301.80361/readings-for-march-31/JJ%20Thomson%20-%20Killing-%20Letting%20Die-%20and%20the%20Trolley%20Problem.pdf) by Judith Jarvis Thomson} \notes{The presentation for each paper will be in the form of an 800 word summary that captures the main message of the work and sets it against the context of the wider discussion we will set up in this first session.} \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. } \thanks \references \ No newline at end of file From a3edc914db2168e4fec6ceccb104cc6309bd313a Mon Sep 17 00:00:00 2001 From: Neil Lawrence Date: Mon, 23 Jan 2023 17:31:46 +0000 Subject: [PATCH 08/10] Update r255 with video and more notes --- ...as-variance-and-fairness-stochasticity-in-decision-making.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md index 11fbbf5..dc0bb59 100644 --- a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md +++ b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md @@ -1 +1 @@ ---- title: "Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" topic: 7 year: 2023 youtube: lLvzTJtYeU8 blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md featured_image: slides/diagrams/ml/bias-variance008.png transition: None --- \notes{\subsection{Overview}} \notes{In this topic we will explore the relationship between the bias-variance dilemma and real world decisions. The opening session will set the context, then we will read the following four papers.} \notes{1. [Neural Networks and the Bias/Variance Dilemma](https://www.dam.brown.edu/people/documents/bias-variance.pdf) by Stuart Geman, Elie Bienenstock, and René Doursat (the technical material of importance is in Section 3). 2. [Probabilistic electrol methods, representative proability and Maximum Entropy](https://www.votingmatters.org.uk/ISSUE26/I26P3.pdf) by Roger Sewell, David MacKay and Ian McLean 3. [Habitual Ethics? Chapter 7](https://library.oapen.org/bitstream/handle/20.500.12657/58884/9781509920433.pdf) by Sylvie Delacroix 4. [Killing, Letting Die and the Trolley Problem](https://learning.hccs.edu/faculty/david.poston/phil1301.80361/readings-for-march-31/JJ%20Thomson%20-%20Killing-%20Letting%20Die-%20and%20the%20Trolley%20Problem.pdf) by Judith Jarvis Thomson} \notes{The presentation for each paper will be in the form of an 800 word summary that captures the main message of the work and sets it against the context of the wider discussion we will set up in this first session.} \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. } \thanks \references \ No newline at end of file +--- title: "Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" topic: 7 year: 2023 youtube: lLvzTJtYeU8 blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md featured_image: slides/diagrams/ml/bias-variance008.png transition: None layout: topic --- \notes{\subsection{Overview}} \notes{In this topic we will explore the relationship between the bias-variance dilemma and real world decisions. The opening session will set the context, then we will read the following four papers.} \notes{1. [Neural Networks and the Bias/Variance Dilemma](https://www.dam.brown.edu/people/documents/bias-variance.pdf) by Stuart Geman, Elie Bienenstock, and René Doursat (the technical material of importance is in Section 3). 2. [Probabilistic electrol methods, representative proability and Maximum Entropy](https://www.votingmatters.org.uk/ISSUE26/I26P3.pdf) by Roger Sewell, David MacKay and Ian McLean 3. [Habitual Ethics? Chapter 7](https://library.oapen.org/bitstream/handle/20.500.12657/58884/9781509920433.pdf) by Sylvie Delacroix 4. [Killing, Letting Die and the Trolley Problem](https://learning.hccs.edu/faculty/david.poston/phil1301.80361/readings-for-march-31/JJ%20Thomson%20-%20Killing-%20Letting%20Die-%20and%20the%20Trolley%20Problem.pdf) by Judith Jarvis Thomson} \notes{The presentation for each paper will be in the form of an 800 word summary that captures the main message of the work and sets it against the context of the wider discussion we will set up in this first session.} \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. } \thanks \references \ No newline at end of file From 2d62803e9d1ca2515587077b00eeef26cfb2b53e Mon Sep 17 00:00:00 2001 From: Neil Lawrence Date: Mon, 23 Jan 2023 17:52:58 +0000 Subject: [PATCH 09/10] Fix problems in header of file. --- _r255/_config.yml | 2 +- ...irness-stochasticity-in-decision-making.md | 88 ++++++++++++++++++- 2 files changed, 88 insertions(+), 2 deletions(-) diff --git a/_r255/_config.yml b/_r255/_config.yml index cc0dee1..3811625 100755 --- a/_r255/_config.yml +++ b/_r255/_config.yml @@ -6,7 +6,7 @@ author: twitter: lawrennd orcid: 0000-0001-9258-1030 url: http://inverseprobability.com -layout: lecture +layout: topic venue: GN06, William Gates Building postsdir: ../../../mlatcl/r255/_topics/ slidesdir: ../../../mlatcl/r255/slides/ diff --git a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md index dc0bb59..7a3a3a9 100644 --- a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md +++ b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md @@ -1 +1,87 @@ ---- title: "Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" topic: 7 year: 2023 youtube: lLvzTJtYeU8 blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md featured_image: slides/diagrams/ml/bias-variance008.png transition: None layout: topic --- \notes{\subsection{Overview}} \notes{In this topic we will explore the relationship between the bias-variance dilemma and real world decisions. The opening session will set the context, then we will read the following four papers.} \notes{1. [Neural Networks and the Bias/Variance Dilemma](https://www.dam.brown.edu/people/documents/bias-variance.pdf) by Stuart Geman, Elie Bienenstock, and René Doursat (the technical material of importance is in Section 3). 2. [Probabilistic electrol methods, representative proability and Maximum Entropy](https://www.votingmatters.org.uk/ISSUE26/I26P3.pdf) by Roger Sewell, David MacKay and Ian McLean 3. [Habitual Ethics? Chapter 7](https://library.oapen.org/bitstream/handle/20.500.12657/58884/9781509920433.pdf) by Sylvie Delacroix 4. [Killing, Letting Die and the Trolley Problem](https://learning.hccs.edu/faculty/david.poston/phil1301.80361/readings-for-march-31/JJ%20Thomson%20-%20Killing-%20Letting%20Die-%20and%20the%20Trolley%20Problem.pdf) by Judith Jarvis Thomson} \notes{The presentation for each paper will be in the form of an 800 word summary that captures the main message of the work and sets it against the context of the wider discussion we will set up in this first session.} \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. } \thanks \references \ No newline at end of file +--- +title: 'Bias, Variance and Fairness: Stochasticity in Decision Making' +date: 2023-01-20 +time: '12:30' +start: '13:00' +end: '15:00' +layout: topic +topic: 7 +year: 2023 +youtube: lLvzTJtYeU8 +blog: 2017-11-15-decision-making.md +blog1: 2018-02-06-natural-and-artificial-intelligence.md +blog2: 2015-12-04-what-kind-of-ai.md +featured_image: slides/diagrams/ml/bias-variance008.png +transition: None +--- + + +\notes{\subsection{Overview}} + +\notes{In this topic we will explore the relationship between the bias-variance dilemma and real world decisions. The opening session will set the context, then we will read the following four papers.} + +\notes{1. [Neural Networks and the Bias/Variance Dilemma](https://www.dam.brown.edu/people/documents/bias-variance.pdf) by Stuart Geman, Elie Bienenstock, and René Doursat (the technical material of importance is in Section 3). +2. [Probabilistic electrol methods, representative proability and Maximum Entropy](https://www.votingmatters.org.uk/ISSUE26/I26P3.pdf) by Roger Sewell, David MacKay and Ian McLean +3. [Habitual Ethics? Chapter 7](https://library.oapen.org/bitstream/handle/20.500.12657/58884/9781509920433.pdf) by Sylvie Delacroix +4. [Killing, Letting Die and the Trolley Problem](https://learning.hccs.edu/faculty/david.poston/phil1301.80361/readings-for-march-31/JJ%20Thomson%20-%20Killing-%20Letting%20Die-%20and%20the%20Trolley%20Problem.pdf) by Judith Jarvis Thomson} + +\notes{The presentation for each paper will be in the form of an 800 word summary that captures the main message of the work and sets it against the context of the wider discussion we will set up in this first session.} + +\include{_philosophy/includes/utilitarianism.md} +\include{_philosophy/includes/utility-utilitarianism.md} +\include{_philosophy/includes/uncertainty-and-absolutism.md} +\include{_ml/includes/what-is-ml-2.md} +\include{_philosophy/includes/bias-towards-variance.md} +\include{_philosophy/includes/bias-variance-in-ml.md} +\include{_philosophy/includes/bias-variance-rational.md} + +\notes{ +\subsection{Fairness in Decision Making}} + +\notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} + +\newslide{} + +\figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} + +\notes{\include{_governance/includes/gdpr-overview.md}} + +\notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} + +\notes{But when considering fairness, it seems that there's two forms that we might consider.} + +\notes{\include{_ai/includes/p-n-fairness.md}} + +\newslide + +\slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} + +\notes{\include{_ai/includes/reflexive-reflective.md}} + +\newslide + +$$\text{reflect} \Longleftrightarrow \text{reflex}$$ + +\newslide + +\newslide{Conclusion} +\slides{ +* Tendency to seek proxies such as *consistency* in decision making. +* Consistent algorithms oversimplifies. +* Society would be more robust if diversity of solutions and opinions are sustained and respected. +} + +\notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } + +\newslide{Links} +\slides{ +* [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) +* [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. +* [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) +* Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) +* [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. +} + +\thanks + +\references From dc369bdd9fc741473bf149a98be28b57d23d968a Mon Sep 17 00:00:00 2001 From: Neil Lawrence Date: Mon, 23 Jan 2023 21:02:19 +0000 Subject: [PATCH 10/10] Update --- ...as-variance-and-fairness-stochasticity-in-decision-making.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md index 11fbbf5..33b162d 100644 --- a/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md +++ b/_r255/bias-variance-and-fairness-stochasticity-in-decision-making.md @@ -1 +1 @@ ---- title: "Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" topic: 7 year: 2023 youtube: lLvzTJtYeU8 blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md featured_image: slides/diagrams/ml/bias-variance008.png transition: None --- \notes{\subsection{Overview}} \notes{In this topic we will explore the relationship between the bias-variance dilemma and real world decisions. The opening session will set the context, then we will read the following four papers.} \notes{1. [Neural Networks and the Bias/Variance Dilemma](https://www.dam.brown.edu/people/documents/bias-variance.pdf) by Stuart Geman, Elie Bienenstock, and René Doursat (the technical material of importance is in Section 3). 2. [Probabilistic electrol methods, representative proability and Maximum Entropy](https://www.votingmatters.org.uk/ISSUE26/I26P3.pdf) by Roger Sewell, David MacKay and Ian McLean 3. [Habitual Ethics? Chapter 7](https://library.oapen.org/bitstream/handle/20.500.12657/58884/9781509920433.pdf) by Sylvie Delacroix 4. [Killing, Letting Die and the Trolley Problem](https://learning.hccs.edu/faculty/david.poston/phil1301.80361/readings-for-march-31/JJ%20Thomson%20-%20Killing-%20Letting%20Die-%20and%20the%20Trolley%20Problem.pdf) by Judith Jarvis Thomson} \notes{The presentation for each paper will be in the form of an 800 word summary that captures the main message of the work and sets it against the context of the wider discussion we will set up in this first session.} \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } \newslide{Links} \slides{ * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. } \thanks \references \ No newline at end of file +--- title: "Bias, Variance and Fairness: Stochasticity in Decision Making" reveal: ipynb: date: 2023-01-20 time: "12:30" start: "13:00" end: "15:00" topic: 7 year: 2023 youtube: lLvzTJtYeU8 blog: 2017-11-15-decision-making.md blog1: 2018-02-06-natural-and-artificial-intelligence.md blog2: 2015-12-04-what-kind-of-ai.md featured_image: slides/diagrams/ml/bias-variance008.png transition: None --- \notes{\subsection{Overview}} \notes{In this topic we will explore the relationship between the bias-variance dilemma and real world decisions. The opening session will set the context, then we will read the following four papers.} \notes{1. [Neural Networks and the Bias/Variance Dilemma](https://www.dam.brown.edu/people/documents/bias-variance.pdf) by Stuart Geman, Elie Bienenstock, and René Doursat (the technical material of importance is in Section 3). 2. [Probabilistic electrol methods, representative proability and Maximum Entropy](https://www.votingmatters.org.uk/ISSUE26/I26P3.pdf) by Roger Sewell, David MacKay and Ian McLean 3. [Habitual Ethics? Chapter 7](https://library.oapen.org/bitstream/handle/20.500.12657/58884/9781509920433.pdf) by Sylvie Delacroix 4. [Killing, Letting Die and the Trolley Problem](https://learning.hccs.edu/faculty/david.poston/phil1301.80361/readings-for-march-31/JJ%20Thomson%20-%20Killing-%20Letting%20Die-%20and%20the%20Trolley%20Problem.pdf) by Judith Jarvis Thomson} \notes{The presentation for each paper will be in the form of an 800 word summary that captures the main message of the work and sets it against the context of the wider discussion we will set up in this first session.} \include{_philosophy/includes/utilitarianism.md} \include{_philosophy/includes/utility-utilitarianism.md} \include{_philosophy/includes/uncertainty-and-absolutism.md} \include{_ml/includes/what-is-ml-2.md} \include{_philosophy/includes/bias-towards-variance.md} \include{_philosophy/includes/bias-variance-in-ml.md} \include{_philosophy/includes/bias-variance-rational.md} \notes{ \subsection{Fairness in Decision Making}} \notes{As a more general example, let's consider fairness in decision making. Computers make decisions on the basis of our data, how can we have confidence in those decisions?} \newslide{} \figure{\includepng{\diagramsDir/data-science/convention-108-coe}{70%}}{The convention for the protection of individuals with regard to the processing of personal data was opened for signature on 28th January 1981. It was the first legally binding international instrument in the field of data protection.}{convention-108-coe} \notes{\include{_governance/includes/gdpr-overview.md}} \notes{The GDPR gives us some indications of the aspects we might consider when judging whether or not a decision is "fair".} \notes{But when considering fairness, it seems that there's two forms that we might consider.} \notes{\include{_ai/includes/p-n-fairness.md}} \newslide \slides{\figure{\includediagramclass{\diagramsDir/ai/n-p-fairness}{80%}}{We seem to have two different aspects to fairness, which in practice can be in tension.}{n-p-fairness}} \notes{\include{_ai/includes/reflexive-reflective.md}} \newslide $$\text{reflect} \Longleftrightarrow \text{reflex}$$ \newslide \newslide{Conclusion} \slides{ * Tendency to seek proxies such as *consistency* in decision making. * Consistent algorithms oversimplifies. * Society would be more robust if diversity of solutions and opinions are sustained and respected. } \notes{Simplistic interpretations of utility theory are misleading about the real decisions we face, and similarly for the machines we design. These simplistic perspctives have also led to a tendency to seek proxies for a notion of "correct" decision making such as consistency. However, in practice uncertainty means that our decisions will often be incorrect. Faced with this incorrectness. Once we accept errors will be made, we can see that making consistent errors is likely to be more harmful than when those errors are inconsistent. } \subsection{Links} * [This paper on decision making and diversity](http://inverseprobability.com/2017/11/15/decision-making) * [This post on System Zero](http://inverseprobability.com/2015/12/04/what-kind-of-ai/) and [This post on the Mechanistic Fallacy](http://inverseprobability.com/2015/11/09/artificial-stupidity/) relate to the ideas in this talk. * [This post on natural vs artificial systems](http://inverseprobability.com/2018/02/06/natural-and-artificial-intelligence) * Articles in the Guardian are available from my [Guardian Profile Page](http://www.theguardian.com/profile/neil-lawrence) * [My blog](http://inverseprobability.com/blog.html) has articles relating to this area. \thanks \references \ No newline at end of file