Many people assume no connection between the two

• Data science is about math/stats and computer science

• Ethics is about social science and philosophy

• The reality is that they are completely intertwined

[S]tudies and evaluates moral problems related to data (including generation, recording, curation, processing, dissemination, sharing and use), algorithms (including artificial intelligence, artificial agents, machine learning and robots) and corresponding practices (including responsible innovation, programming, hacking and professional codes), in order to formulate and support morally good solutions

– Floridi & Taddeo (2016)

Consider this statement:

Data are inherently objective, but people are not

Consider this statement:

Data are inherently objective, but people are not.

Do you agree? Why or why not?

• Findable

• Accessible

• Interoperable

• Reusable

• FAIR principles largely focus on aspects that will facilitate increased data sharing

• They largely ignore power differentials and historical contexts

• This creates a tension for Indigenous Peoples who are asserting greater control over the application and use of Indigenous data & knowledge for collective benefit

• Collective benefit

• Authority to control

• Responsibility

• Ethics

• Variation in reporting rates

• Variation in police attention

• Variation in rates of enforcement

• Collecting a random sample is difficult

• ML will reproduce the biases in the data used to train it

• We need to think about what’s missing from the training data and what types of biases the data encode

• We need to be aware of the consequences of reinforcing those biases

Motivation

• Provide ecology faculty job seekers info about the job market

• Evaluate whether current hiring is diverse & equitable

Approach

• Positions & applicants for 2016-2018 scraped from ecoevojobs.net

• Determine who was hired into those jobs

• Infer genders from names & pictures

Do you see any ethical problems with this approach?

It is important to identify any systemic gender disparities in the ecology faculty job market, and to identify their causes so that the disparities can be remedied. Inferring a gender binary from a person’s name, as I did, is standard practice in research on gender disparities in many areas, including in ecology and allied fields. This approach performs well. In cases of ambiguity, standard practice is to resolve the ambiguity where it is feasible to do so, by consulting publicly-available photographs and pronoun use in social media profiles. I did so.

Using a gender binary, and inferring the genders of some new hires from their names, is an imperfect approach. Gender is not binary, and ecologists who do not identify as men or women are our colleagues.

Are the data valid for their intended use?

Have we identified & minimized any bias in the data or in the model?

Have we identified & minimized any bias in the data scientist?

Is the analysis transparent and reproducible?

What are the likely misinterpretations of the results and what can be done to prevent them?

• Lawfulness

• Competence

• Dealing with data

• Algorithms and models

• Transparency, objectivity & truth

• Working alone and with others