Let’s revisit

AI Guild Values and Code of Conduct

What AI Guild membership means

This September, we would like to highlight the AI Guild Values & Code of Conduct.

The original post by Marcel Ackermann can be found here:

AI Guild Values & Code of Conduct

“In the AI Guild we care about good and ethical behaviour. In a long community process, we formalized this intention in the following document. Iterating from the founding letter, we refined our values, made them actionable in the form of the code of conduct with expected and unacceptable behaviour and give implementation guidelines. Living the values and adherence to the code of conduct is binding for our members, and we encourage everyone working in the area of AI to consider doing the same. This is a living document, feedback is highly appreciated.

Values

Code of Conduct
1. Ethical AI
2. Diversity
3. Communication
4. Best idea wins
5. Responsibility

Implementation
1. Consequences of Unacceptable Behaviour
2. If You Witness or Are Subject to Unacceptable Behaviour
3. Addressing Grievances
4. Conflicts
5. Conflict of Interest
6. Scope

Values

AI Guild activities are carried out in alignment with our values. Every AI Guild member accepts these values, is fully aware of them, and committed to act in alignment.

1. Ethical AI: We commit for the ethical application of AI that contributes to the public good and the progress of humanity
2. Diversity: We stand for diversity, and are providing a friendly, safe and welcoming environment for everyone regardless of gender, sexual orientation, ability, ethnicity, nationality and socioeconomic status
3. Communication: We treat each other and our partners with respect, fairness, empathy and kindness, and communicate transparently and honestly
4. Best idea wins: We follow a best idea wins approach informed by scientific evidence and governed by democratic processes
5. Responsibility: We take responsibility for every action we take

Code of Conduct

This Code of Conduct outlines our expectations for all those who participate in our community, as well as the consequences for unacceptable behaviour.

This document is directly linked with our values. For each value we provide a list of positive examples and negative examples; behaviour that’s encouraged and behaviour/actions that are clearly not acceptable.

Harmful or unethical inactions are as bad as harmful or unethical actions.

We invite all those who participate in our events to help us create a safe and positive experience for everyone.

1. Ethical AI

Expected Behaviour

The ethical application of AI is integral to the AI Guild’s mission. In this section we provide a general overview, but a comprehensive detailed perspective is beyond the scope of this document.

We advocate adopting AI technologies in existing or new profit or non-profit, governmental or non-governmental organizations with high awareness of data privacy policies and the possible impact of the technology on humanity.

We take data privacy seriously. To lead in this regard, we transparently communicate to our customers or any other data sources when and how exactly the data collected from them directly or indirectly via their interaction will or might be used. We only use data that has been collected consensually. Privacy is a basic right that enables agency, trust and freedom.

We take our own “AI Hippocratic Oath” to strive to do no harm, accelerating AI adoption in a way that positively influences living beings and the environment. An example of ethical AI adoption might be integrating system tests into new services which help to eliminate side effects of the technology such as unintentional biases around sex, age or skin colour of the developed algorithms.

We regard explainability and accountability as an important component of the ethical application of AI. The processes involved in an AI system should be as transparent as possible, its purpose and capability communicated openly. Decisions made by AI systems should be auditable and should not supplant human responsibility. Parties affected by them should have the possibility to lodge complaints and seek redress. We support and are committed to open-sourcing software, except when there is a risk of malicious use or it contradicts our ethical guidelines.

We the AI Guild appreciate any usage of AI which helps businesses or non-profit organizations to improve their services or make their everyday tasks easier. Good examples of AI adoption include:

• Health Care: improving availability and quality of diagnosis, personalized medicine.
• Communication: Machine Translation, Text2Speech, Speech2Text to improve accessibility of information and allow people to communicate across language barriers.
• Education: increasing the availability and quality of education measures and educational tools through e.g. reduced cost, increased interactiveness, personalisation.
• Mobility: Self-driving cars to reduce accidents and increase comfort.
• Robotics: replace dangerous or hard labour, mundane tasks (cleaning).
• Smart Personal Assistants: save time, increase productivity and automate monotonous tasks.
• Wildlife conservation: automatically photograph, classify and count wild animals.

We take into consideration the social impact of the systems we build and design metrics and objective functions with consideration for properties like fairness and privacy.

Unacceptable Behaviour

AI gives a lot of power into the hands of anyone who is familiar with the technology, has sufficient knowledge about it and knows how to use it. Power in AI Guild understanding means also responsibility. However, in some cases it might happen that the AI adoption will be implemented in a way, where the actors are not aware of the possible impacts of the technology or it’s side effects. We do generally discourage anyone to initiate an AI adoption without gaining sufficient knowledge on the topic or getting experienced AI professionals on board. Examples of AI adoption failures are:

• Health Care: relying on the outcome of the machinery diagnosis without questioning the results, violating patient privacy and anonymity.
• Energy Sector: building algorithms (intentional or unintentional) which might lead to a blackout or a system overload or an accident.
• Financial Sector: consultancy services, where decision on whether a person might get a credit or not is based on sex, age or skin colour instead of more reliable and objective backgrounds.
• Politics: not giving enough information to the system, or not covering all possible candidates. This might lead to influencing of the outcome. Here it doesn’t matter whether it was manipulated, or the failure is coming out of the insufficient knowledge of the developers. The impact might be catastrophic.
• Judicial system: bias against minorities.
• Face recognition for surveillance, identifying and tracking individuals with AI.
• Personalized news feeds that manipulate users with fake news.
• Covert AI systems: not letting the user know that she/he is dealing with an AI.
• AI-enabled citizen scoring in violation of fundamental rights.
• Lethal autonomous weapon systems.
• Cyberwar, AI viruses/malicious software.

The impact of the wrong AI adoption might lead to irreversible harm to living beings and the environment intentionally or unintentionally. The AI Guild therefore sees itself as an organisation which acts in a way that mitigates these risks and encourages everyone to be aware of them as well as taking proactive actions towards ethical AI.

2. Diversity

Expected Behaviour

We believe that having people with diverse backgrounds and opinions fosters progress. Besides that, we value every opinion regardless of the academic degree, gender, status or nationality of the person. Some examples of practices that promote diversity are:

• Keep the focus on skills, abilities and ideas of members and partners, not on other socially identifying features.
• Build and maintain a culture of inclusiveness, make everyone feel welcome.
• Appreciate everyone’s contributions, bring up accomplishments, provide chances to grow skills, listen, support each other.
• Share valuable information about projects and opportunities through channels that are accessible to everyone and not through channels that favour people with certain (majority) attributes.
• Establish indicators of diversity and keep track of them.
• Promote the AI Guild to underrepresented groups.
• Ensure diversity in the decision-making, not only in membership in general.

Unacceptable Behaviour

We don’t tolerate intentional discrimination and we also work on rectifying natural social dynamics that result in unequal access to skills and opportunities for different groups. Some examples of the practices that we consider harmful are:

• Preferential treatment based on race, gender or nationality.
• Recruitment practices limited to certain elite schools.
• Pre-judging people based on membership of social groups.
• Speech or behaviours that are offensive to members of certain groups, particularly when they indicated it.
• Conducting meetings in venues that inconvenience and exclude some participants.
• We have members from conflicting countries. Some of them are in the pre/post or ever war state some of them are in the stage of the so-called frozen conflict. AI Guild is not a place to take interstate conflicts to the personal level.
• The AI Guild is not a place of political polarization or tribalism and so discussions of politics should focus on topics relevant to the AI Guild and its activities.

3. Communication

Expected Behaviour

We make sure that the efforts of every member are appreciated. We attempt to find a collabourative way of working together in any disagreements. We praise each other vocally or in a written form and by doing this we give positive feedback for the good things any of us do (as a recommendation: keep the ratio of positive to negative interactions above 5:1).

Feedback given to each other should be honest and constructive. When giving feedback, we do not question someone’s ability, the time period of them belonging to the AI Guild, or any other factors which do not relate directly to the action or idea being discussed.

We have a policy of flat hierarchy and anyone’s efforts and ideas are welcomed and will be evaluated according to Paragraph 4.

Good examples here are:

• When someone has put an effort into moving things forward within AI Guild scope, before saying anything, we give words of appreciation.
• By criticizing someone’s work, we give constructive feedback and possible solutions which from the point of view of the criticizer might provide better results.

Unacceptable Behaviour

We don’t tolerate physical, verbal, psychological, sexual or social violence, or violence of any other kind, in our interactions. We do not withhold information that could be relevant to the decision-making processes of the AI Guild, its governing bodies and individual members except if justified through declared conflicts of interest. Similarly, it is unacceptable to hide information that would allow us to evaluate our actions’ outcomes, both when trying to recreate successes and avoid failures. We don’t do mansplaining.

We don’t use arguments like:

• “You joined the AI Guild yesterday, so your opinion is not relevant.”
• “You do not have a degree in this field, therefore your opinion is less valuable than someone with a PhD or longer experience.”
• “Didn’t anybody tell you about the meeting yesterday? Too bad, now you can’t join our discussion.”

4. Best idea wins

Expected Behaviour

In case we have several solutions to one problem, we compare ideas on their merits without considering who brought them up. Each solution will be presented to the members of the AI guild (or committee). This presentation of the solutions (short summary or oral presentation) should answer the questions why, what, how, and when (or how long). This includes:

• the overall approach
• Intrinsic pros and cons of the approaches, such as “we can expect an intermediate/mediocre result but within a short period of time” or “we can expect high output values, but the work requires a team of 5 people working on it for a year”
• the motivation for adopting the solution, “what’s the opportunity?”, or inversely the bad examples for not having adopted it so far/ consequences of not adopting the suggested solution
• the metrics: ethical reasoning, alignment with output values, budget, timing, other relevant indicators (or realistic estimates when certainty is not achievable)

We generally adhere to scientific methods and processes, and obtain as much evidence as possible. The different approaches will be compared based on those parameters mentioned above. If one solution does not win over the other, the decision will be put to a vote within our community.

Unacceptable Behaviour

Questioning an idea based on the person who brought it up. Especially if it’s based on their ability, gender, age, period belonging to the AI Guild, nationality, beliefs or personality.

Arguing with or shutting down options based on personal evidence/emotional attachment, appealing to authority, appeal to common sense and other fallacies (https://en.wikipedia.org/wiki/List_of_fallacies).

5. Responsibility

Expected Behaviour

We are aware that our community and the actions of each member might have bigger or smaller impacts on the outside world. Therefore, we expect members, governing bodies and the AI Guild as a whole to take full responsibility for their actions regardless of whether an outcome was positive or negative.

Taking responsibility may take different shapes, e.g. personal and indirect responsibility on choices/decisions made. The AI Guild guarantees that its decision-making processes are transparent and in accordance with Sec. 4. Individual members that partake in decision-making processes of the AI Guild are expected to voice their opinions and will be given an opportunity to dissent. Individual members are expected to reflect on the values set forth in this Code of Conduct when considering which actions to take. Actual and perceived conflicts of interest are to be declared as explained in a later section.

After actions have been carried out, taking responsibility entails evaluating the outcome honestly and transparently. This requires reflecting both on mistakes and successes made along the way and developing strategies to avoid the former and recreate the latter in the future. Thereby, it is important to differentiate honest mistakes, oversights, irresponsible actions and malicious actions. Furthermore, an honest evaluation means being aware that a positive outcome might have been achieved by unjust, irresponsible or reckless means.

Lastly, taking responsibility entails a commitment to continuous improvement.

Unacceptable Behaviour

We do not retroactively diminish or exaggerate our involvement in actions taken to avoid consequences or obtain undo appraisal. Similarly, we do not employ after-the-fact criticisms along the lines of “I knew this would happen/I could have told you so” but instead voice our opinions, possible reservations we might have and our inclination to dissent during our decision-making processes.

We do not present irresponsible actions as unintentional mistakes.

We do not repeat strategies that were successful by sheer coincidence and we do not follow an ends-justify-the-means approach.

Implementation

1. Consequences of Unacceptable Behaviour

Unacceptable behaviour from any community member, including sponsors and those with decision-making authority, will not be tolerated. Anyone asked to stop unacceptable behaviour is expected to comply immediately.

If a community member engages in unacceptable behaviour, the community organizers may take any action they deem appropriate, up to and including a temporary ban or permanent expulsion from the community without warning.

2. If You Witness or Are Subject to Unacceptable Behaviour

If you are subject to or witness unacceptable behaviour, or have any other concerns, please notify someone from the organizer or ethics group as soon as possible. Additionally, community organizers are available to help community members engage with local law enforcement or to otherwise help those experiencing unacceptable behaviour feel safe. In the context of in-person events, organizers will also provide escorts as desired by the person experiencing distress.

3. Addressing Grievances

If you feel you have been falsely or unfairly accused of violating this Code of Conduct, you should notify one of the event organizers with a concise description of your grievance. Your grievance will be handled in accordance with our existing governing policies.

4. Conflicts

We encourage constructive disagreements, and trust that in most cases they can be resolved by the members themselves. Should you require assistance in conflict resolution, our mediators are there to help. We don’t tolerate physical, verbal, psychological, sexual or social violence and will treat it as a violation of this Code of Conduct.

5. Conflict of Interest

When acting on behalf of the AI Guild, such as being a member of a committee or an event organizer, you need to declare any real or perceived conflict of interest. Reach out to one of our mediators for advice on what is and is not a conflict of interest and how it should be handled. Please refer to the EU guidelines that we model our approach on.

6. Scope

We expect all community participants (contributors, paid or otherwise; sponsors; and other stakeholders) to abide by this Code of Conduct in all community venues — online and in-person — as well as in all one-on-one communications pertaining to community business.”