AlgorithmWatch est une organisation non-gouvernementale à but non lucratif. Nous analysons les systèmes algorithmiques ayant une influence sur la société, utilisés soit pour prédire ou encourager des comportements, soit pour prendre des décisions automatiquement. Nos activités sont la recherche d’information, l’analyse et la promotion de politiques publiques qui bénéficient à la société dans son ensemble.

Nous avons formulé notre charte en tant que Vision, Mission & Valeurs dans le texte ci-dessous. Le texte n'est actuellement disponible qu'en anglais. Nous publierons prochainement une version française. La charte complète le Manifeste ADM qui guide notre travail depuis la création d'AlgorithmWatch. Vous trouverez le manifeste (en français) plus bas sur cette page.

Vision, Mission & Valeurs

Our vision

We envision a world where technology in general, and algorithmic systems in particular, are used to protect and bnefit individuals and to make societies more just, democratic, inclusive and sustainable, be it with regard to ethnicity, gender and sexual orientation, abilities, age, wealth, class, or resource consumption.

Our mission

To ensure that the use of algorithmic systems benefits the many, not the few, we advocate for algorithmic systems that protect democracy and the rule of law instead of authoritarianism, freedom instead of surveillance, human rights instead of dehumanization, autonomy instead of power asymmetries, justice and equality instead of favoritism and disparity, and a sustainable instead of an exploitative way of living.

To achieve this mission, we

• develop and use tools to unpack and analyze the use of algorithmic decision-making processes and their effects on humans and societies;
• explain the characteristics of and power structures behind these complex algorithmic decision-making processes. We point out their risks, conflicts and possible harms to a general public and decision-makers and help them distinguish between applications grounded in scientifically solid, ethically responsible research and snake oil;
• develop effective frameworks governing the use of algorithmic systems, based on our research and scientific evidence, by ensuring transparency, individual and democratic control, and accountability;
• publicly call out misuses and negative impacts of algorithmic systems on human rights and the public good, and campaign for effective transparency and accountability mechanisms. We advocate for the recognition and inclusion of diverse perspectives into all aspects of the use of and the oversight over algorithmic systems, including decisions about whether their application is justified for a given purpose;
• build a community of people and organizations from different backgrounds, cultures and disciplines to collaborate towards these goals;
• promote the broader debate on whether there should be limits to the use of computation-based approaches to human problems, and on how they irreducibly differ from non-computation-based approaches.

Our values

At AlgorithmWatch, we practice what we expect from others: respect, fairness, solidarity, integrity, inclusivity, diversity, transparency, accountability. We live a culture of both rigor and forgiveness. We give our best and accept that mistakes are a necessary part of learning, both individually and as an organization. We believe in encouragement, not degradation. We stand by each other.

Our positions and demands are based on evidence gained by professional analysis and high standards of research. We are politically non-partisan and non-sectarian. We call out nonsense no matter where it comes from, stand up to power, and are ourselves open to being challenged from anyone with a reasoned argument. We engage with anyone who exhibits genuine goodwill towards solving problems, be it civil society, governments, public authorities, academia, or private companies.

We are convinced that civil society and individuals and groups affected by algorithmic systems must be a key voice in this debate, and we strive to be a convincing and respectful representative of the public interest.

We only accept funding that does not compromise our ability to address issues freely, thoroughly and objectively. We produce and publish accurate and timely reports of our activities and funding.

As a team, we acknowledge people as individuals with dignity, not as human resources whose value can be quantified and measured by performance metrics. We understand that being a diverse and inclusive team is a continuous process. To make it a successful one, we have to be willing to listen and learn, to see and question privilege and act upon the results. We strive for balanced and diverse representation on our governing bodies and our team.

We don't take ourselves more seriously than necessary. We believe that a day without laughing together and about ourselves is a wasted day.

---

Notre manifeste

La prise de décision algorithmique (algorithmic decision-making, ADM) est une réalité aujourd'hui ; ce sera une réalité encore plus importante demain. Tout en étant prometteur par de nombreux aspects, l’ADM porte en elle d'énormes dangers. Le fait que la plupart des systèmes d’ADM soient des « boîtes noires » n’est pas une fatalité.

1. Un système d’ADM n'est jamais neutre.
2. Le créateur ou la créatrice d'un système d’ADM est responsable de ses effets. Par ailleurs, un système d’ADM n'est pas uniquement créé par son concepteur ou sa conceptrice mais englobe des éléments créés par d’autres.
3. Un système d’ADM doit être intelligible, afin de pouvoir rendre des comptes aux institutions qui permettent la démocratie.
4. Les sociétés démocratiques doivent rendre les systèmes d'ADM intelligibles en utilisant à la fois des technologies, des lois et des instances de contrôle.
5. Nous devons pouvoir décider des libertés qu’il ou elle délègue à un système d’ADM.

Pour nous, un processus de prise de décision algorithmique implique les étapes suivantes:

• Conception de procédures de recueil de données,
• Recueil de données,
• Conception d’algorithmes pour
• analyser les données,
• interpréter les résultats de cette analyse sur la base d'un modèle créé par des humains,
• agir automatiquement, de quelque manière que ce soit, à partir de ces interprétations.