Governance and Ethics Assigment

A short-paper (2-3 pages, excluding references) should be written in groups of 3. You should 

1. Introduce the case
2. Give examples if you find them
3. Make an analysis based on the lecture: 

We want you to focus on at least three common themes in governance: 

• Accountability – how should responsibilities best be distributed? Basically, if something breaks or goes wrong, who is to blame? Who should ensure that nothing goes wrong?
• Fairness – what would be a fair approach in terms of law, ethics or social norms in this case? So, in short, what is fair here? 
• Transparency – what level or type of transparency, and for whom, would be fitting for this case? So, for example, who should understand how a prediction works – a customer, a doctor, a handling officer, the developer, etc? They will all have different expectations and dependencies. 

To clarify , all of these aspects can be further pointed to in terms of 

• roles, as in whom are involved or affected; persons or user of various sorts, companies, authorities, or other stakeholders; 
• context, how is it relevant, if it is?
• stakes, that is, the magnitude of risks

Feel free to expand outside this box, find references for your arguments. Remember, there may not necessarily be one normatively correct answer for what is supposed to be fair, but several different possible approaches or ways to balance between different interests. The norms – both legal and social – could also be anticipated to change over time. If so, point to that. 

You will be randomly allocated one of the following cases to analyse in your writing group. The examination will be done through a written hand-in as well as a brief presentation. Each case will have a corresponding review group that will provides feedback based on the written review of the case-group. 

Cases: 

1. Image-recognition to aid in mammography screening

In Sweden all women between 40 and 74 years-old are offered free screening for breast cancer. This is in general done by radiologists assessing mammography pictures to detect any riskful anomalies that need to be followed up. In practice, this means that a handful of radiologists look at thousands of pictures per year in each region. In parallel, image-recognition has proven to become fairly good at this screening task, and the question now is under what circumstances such an aid can be used to help the screening programs. 

2. Social robots as domestic carers for people with disabilities

Social robots are robots that can interact with humans, as well as assist them. In the context of care, this can be by monitoring the user’s health, whilst also playing games. Growing research is pointing towards social robots being able to care in homes for the users.

3. Predictions and automation in the distribution of welfare in public institutions

State level authorities in many countries see great potential in creating more efficient and scalable handling of citizen welfare. The distribution of welfare in social insurance, the allocation of resources in public employment services or the fraud detection amongst tax authorities all see benefits in a machine learning-based approach. 

4. Decision-support for chest pain assessment in emergency care

Experiencing pain in the chest is a fairly common reason for people to go to the emergency care. A majority of these cases are however luckily not of a very serious character, and only a small minority is serious. To tell the difference quickly is however hard, why many patients that do not need further care are hospitalised and consume resources that could be better used for those in more dire need. Therefore, machine-learning based decision support systems are being developed, based on the early tests that can be made on patients in the emergency care, in combination with other information already available.

5. “Smart cameras” in public places

Many see benefits in using image-recognition in so-called smart cameras in public environments. This could include facial recognition, and police interests in making arrests or looking for missing people, but also ways to create swift ticketing outside sports arenas or at railway stations.

6. Dead people’s data used to train individual chatbots

A recent Microsoft patent describes the functionality of creating a chatbot based on data from a specific individual – “past or present”, and mentions friends, family members or acquaintances. Given that data protection is weak or absent for deceased people, it is possible to consider using data from dead people to develop chatbots. The data can regard voice, writing behaviour, photos etc.