knowledge base update

Author: jfparie

content/.DS_Store

@@ -33,7 +33,7 @@ about_AA:
         [op-eds](/knowledge-platform/knowledge-base/).
 team:
   title: Synthetic data generation cohort
-  content: Part-time team working on synthetic data generation
+  content: Part-time team working on synthetic data generation for AI bias testing
   icon: fas fa-table
   button_text: Synthetic data generation
   id: SDG-team
@@ -61,9 +61,9 @@ team:
         Research Scientist, Spotify
 team1:
   title: Bias detection tool cohort
-  content: Part-time team working on the bias detection tool
+  content: Part-time team working on unsupervised the bias detection tool
   icon: fas fa-search
-  button_text: Our bias detection tool
+  button_text: Bias detection tool
   id: bdt
   button_link: /technical-tools/bdt/
   team_members:
@@ -97,7 +97,7 @@ team2:
       name: Jurriaan Parie
       bio: |
         Director-board member
-    - image: /images/people/VDjwalapersad.png
+    - image: /images/people/VDjwalapersad.jpeg
       name: Vardâyani Djwalapersad
       bio: |
         Project manager Algoprudence

content/english/about/teams.md

@@ -1,12 +1,13 @@
 ---
-title: >-
-  How 'algoprudence' can contribute to responsible use of ML-algorithms
-author: 'Anne Meuwese, Jurriaan Parie & Ariën Voogt'
+title: How 'algoprudence' can contribute to responsible use of ML-algorithms
+subtitle: ''
 image: /images/knowledge_base/NJB-cover.jpg
+author: 'Anne Meuwese, Jurriaan Parie & Ariën Voogt'
 type: featured
 summary: >-
-  By means of two case positions regarding the use of machine learning-driven risk profiling by the municipalities of Rotterdam and Amsterdam, the concept of 'algoprudence' is introduced and explained. 
-subtitle: ''
+  By means of two case positions regarding the use of machine learning-driven
+  risk profiling by the municipalities of Rotterdam and Amsterdam, the concept
+  of 'algoprudence' is introduced and explained.
 ---
 
 Article in journal for Dutch legal scholars #10 https://www.njb.nl/magazines/njb-10-2024/
@@ -15,4 +16,4 @@ Article in journal for Dutch legal scholars #10 https://www.njb.nl/magazines/njb
 
 By means of two case positions regarding the use of machine learning-driven risk profiling by the municipalities of Rotterdam and Amsterdam, the concept of 'algoprudence' is introduced and explained. This new term refers to concrete, case-based, and decentralized judgments regarding the normative decision made to design and deploy algorithms. The article illustrates that the general principles of sound administration (as specified in Dutch Administrative Law) are insufficient to provide concrete standards for these algorithms, the authors argue that algoprudence can serve as a useful addition to and concretization of existing legal frameworks.
 
-{{< pdf_frame articleUrl1="https://drive.google.com/file/d/1V9k4ghq4RJHu_9UMiA8QctkcDX3teQYR/preview" >}}
+{{< pdf_frame articleUrl1="https://drive.google.com/file/d/1fIZ9oRTULNOlgzk6Dwsr3hYujSiRu85n/preview" width_desktop_pdf="6" width_mobile_pdf="12" >}}

content/english/knowledge-platform/knowledge-base/NJB-artikel.md

@@ -1,11 +1,10 @@
 ---
-title: >-
-  Algorithm Audit (white paper) – A new algorithmic regulatory body in The
-  Netherlands
+title: Algorithm Audit (op-ed) – A new algorithmic regulatory body in The Netherlands
+subtitle: ''
+image: /images/knowledge_base/Op-ed-1.png
 author: Algorithm Audit
-image: /images/knowledge_base/white-paper-2.png
 type: regular
-summary: "A new algorithmic regulatory body is institutionalized in The Netherlands. We say: Make it a bulldog \U0001F43A. Not a lap dog \U0001F436."
+summary: "A new algorithmic regulatory body is institutionalized in The Netherlands. We say: Make it a bulldog \U0001F43A, not a lap dog \U0001F436"
 ---
 
 New white paper on algorithmic supervision. We reflect on recently announced plans by the Dutch Minister of Digitalization to institutionalize a national algorithmic watchdog. We say: Make it a bulldog <span style="font-size: 25px;">🐺</span>. Not a lap dog <span style="font-size: 25px;">🐶</span>.
@@ -14,6 +13,6 @@ New white paper on algorithmic supervision. We reflect on recently announced pla
 
 The new regulatory body seems to rely on the principle of ‘market supervises market’. As an effect, commercial firms get a key position in defining wat is ‘fair’ and what not. But, this is a public tasks that ought to be performed in the democratic sight. Read the full white paper (2 pages) below.
 
-<span style="font-size: 14px; font-style:italic">This white paper is published as an [op-ed](https://fd.nl/opinie/1462782/maak-nieuwe-algoritmewaakhond-een-bulldog-in-plaats-van-een-schoothond) in the Dutch newspaper *Het Financieele Dagblad* on December 30th 2022.</span>
+<span style="font-size: 14px; font-style:italic">This white paper is published as an <a href="https://fd.nl/opinie/1462782/maak-nieuwe-algoritmewaakhond-een-bulldog-in-plaats-van-een-schoothond" target="_blank">op-ed</a> in the Dutch newspaper *Het Financieele Dagblad* on December 30th 2022.</span>
 
-{{< pdf_frame title="White-paper" name="white-paper" articleUrl="https://drive.google.com/file/d/18GyjLGhKqzBmYIqWP8Y20BcxVbxyfuLR/preview" >}}
+{{< pdf_frame articleUrl1="https://drive.google.com/file/d/1gZo_1IUFuCmTFZfHW4j7jSk9QzCyGbxN/preview" width_desktop_pdf="6" width_mobile_pdf="12" >}}

…ite_paper_algorithmic_regulatory_body.md …ase/Op-ed_algorithmic_regulatory_body.mdcontent/nederlands/knowledge-platform/knowledge-base/White_paper_algorithmic_regulatory_body.md renamed to content/english/knowledge-platform/knowledge-base/Op-ed_algorithmic_regulatory_body.md content/nederlands/knowledge-platform/knowledge-base/White_paper_algorithmic_regulatory_body.md renamed to content/english/knowledge-platform/knowledge-base/Op-ed_algorithmic_regulatory_body.md

@@ -0,0 +1,19 @@
+---
+title: Algorithm Audit (op-ed) – Human vs machine bias
+subtitle: ''
+image: /images/knowledge_base/Op-ed-3.png
+author: 'Jurriaan Parie, Vardâyani Djwalapersad'
+type: regular
+summary: >-
+  Based on bias testing results from the Municipality of Amsterdam, it is argued
+  that that algorithms can play an important role in mitigating biases
+  originating from humans
+---
+
+Op-ed, as <a href="https://www.parool.nl/columns-opinie/opinie-onderzoek-vooringenomenheid-van-zowel-algoritme-als-ambtenaar~bd69aa5e/" target="_blank">published</a> in Parool on 14-02-2024, arguing that:
+
+* Not only algorithmic-driven processes can have discriminatory effects, but that human-driven process can be severely biased too;
+* They argue therefore that, as a result of the performed bias test by the City of Amsterdam, not only should the explainable boosting ML-model be abandoned, but also the allegedly detected human biases within the processes of the City of Amsterdam should be subject to further investigation;
+* Because more open and transparent research is needed to strengthen human-machine interplay to prevent systemic biases in the digital future.
+
+{{< pdf_frame articleUrl1="https://drive.google.com/file/d/1rtXD82BsXu2qRPlIemOgKYijZMYSrRNT/preview" width_desktop_pdf="6" width_mobile_pdf="12" >}}

content/english/knowledge-platform/knowledge-base/Op-ed_human_vs_machine_bias.md

@@ -1,11 +1,14 @@
 ---
-title: Algorithm Audit (white paper) – Reversing the burden of proof
+title: Algorithm Audit (op-ed) – Reversing the burden of proof
+subtitle: ''
+image: /images/knowledge_base/Op-ed-2.png
 author: Algorithm Audit
-image: /images/knowledge_base/white-paper-1.png
-type: "regular"
-summary: Algorithm Audit's first white paper on the reversal on the burden of proof in the context of (semi-)automated decision-making
-
+type: regular
+summary: >-
+  Algorithm Audit's first white paper on the reversal on the burden of proof in
+  the context of (semi-)automated decision-making
 ---
+
 Algorithm Audit has published its first white paper🥇 Reversing the burden of proof is a promising approach to protect against discrimination in the context of (semi-)automated decision-making (ADM). Yet to make it feasible we need to overcome some hurdles.
 
 This is what we advise policy makers across the EU to facilitate reversing the burden of proof and improve legal protection for (semi-)ADM:
@@ -14,14 +17,12 @@ This is what we advise policy makers across the EU to facilitate reversing the b
 
 <span style="font-size: 25px;">🤳</span>Establish national hotlines where potentially discriminatory (semi-)ADM can be reported and where litigation support can be provided;
 
-<span style="font-size: 25px;">📚</span>Train staff of courts, parliament, ministries, and other authorities in interpreting (semi-)ADM qualitatively. 
+<span style="font-size: 25px;">📚</span>Train staff of courts, parliament, ministries, and other authorities in interpreting (semi-)ADM qualitatively.
 
-<span style="font-size: 14px; font-style:italic">This white paper is published as an [op-ed](https://fd.nl/opinie/1436425/we-moeten-ons-bezinnen-op-het-bestaansrecht-van-algoritmen) in the Dutch newspaper _Het Financieele Dagblad_ on April 13th 2022.</span>
+<span style="font-size: 14px; font-style:italic">This white paper is published as an [op-ed](https://fd.nl/opinie/1436425/we-moeten-ons-bezinnen-op-het-bestaansrecht-van-algoritmen) in the Dutch newspaper *Het Financieele Dagblad* on April 13th 2022.</span>
 
 ###### **Summary**
 
 This document formulates actionable suggestions to improve legal protection for citizens and consumers in the European Union in the context for (semi-)automated decision-making (ADM). The suggestions in this document are linked to an existing concept in EU non- discrimination law: the reversal of the burden of proof.
 
-{{< container_open >}}
-{{< pdf_frame title="White-paper" name="white-paper" articleUrl="https://drive.google.com/file/d/1RHdqoGVgwv-FTv8qC9fAlsVl8eUTcR7s/preview" >}}
-{{< container_close >}}
+{{< pdf_frame articleUrl1="https://drive.google.com/file/d/1gZo_1IUFuCmTFZfHW4j7jSk9QzCyGbxN/preview" width_desktop_pdf="6" width_mobile_pdf="12" >}}

…/White_paper_reversal_burden_of_proof.md …e-base/Op-ed_reversal_burden_of_proof.mdcontent/english/knowledge-platform/knowledge-base/White_paper_reversal_burden_of_proof.md renamed to content/english/knowledge-platform/knowledge-base/Op-ed_reversal_burden_of_proof.md content/english/knowledge-platform/knowledge-base/White_paper_reversal_burden_of_proof.md renamed to content/english/knowledge-platform/knowledge-base/Op-ed_reversal_burden_of_proof.md

@@ -1,15 +1,15 @@
 ---
 title: >-
-  Algorithm Audit (white paper) – Feedback on DSA Delegated Regulation (conducting independent
-  audits)
-author: Algorithm Audit
+  Algorithm Audit (white paper) – Feedback on DSA Delegated Regulation
+  (conducting independent audits)
+subtitle: ''
 image: /images/knowledge_base/white-paper-3.png
+author: Algorithm Audit
 type: regular
 summary: >-
   Plea to include the normative dimension of AI auditing in delegated regulation
   of the Digital Services Act (DSA). Current limitations are illustrated by
   focussing on a recommender systems example
-subtitle: ''
 ---
 
 Feedback to the European Commission on DSA Delegated Regulation – conducting independent audits.
@@ -18,8 +18,4 @@ Feedback to the European Commission on DSA Delegated Regulation – conducting i
 
 In addition to Article 37 of the Digital Services Act (DSA), Delegated Regulation (DR) sets out procedures, methodologies and templates for third-party auditing of Very Large Open Platforms (VLOPs) and Very Large Online Search Engines (VLOSEs). The DR builds upon established sector-specific risk management frameworks to provide procedural guidance for AI audits. However, the regulation lacks provisions to disclose normative methodological choices that underlie AI systems (e.g., recommender systems), which is crucial for evaluating associated risks in a meaningful way (as mandated by DSA Article 34). To illustrate this limitation, we elaborate on methodological crossroads that determine the performance of recommender systems and its downstream risks. We make concrete suggestions how the definition of ‘inherent risk’ (Article 2), audit methodologies of risk assessments (Section IV) and the audit report template (Annex I) set out by the DR should be amended to incorporate normative dimension of AI auditing in a meaningful way. Only if both the technical and normative dimension of AI systems are thoroughly examined, risk assessed under the DSA will empower the European Union and its citizens to determine what public values should to be safeguarded in the digital world.
 
-{{< container_open >}}
-
-{{< pdf_frame >}}
-
-{{< container_close >}}
+{{< pdf_frame articleUrl1="https://drive.google.com/file/d/1v6CApiRsT4vE1e-EXJnHDufk0FyXLHwL/preview" width_desktop_pdf="6" width_mobile_pdf="12" >}}