Photo by @alexander_tsang

AI Safety and Social Data Science

University of Copenhagen and its newly created Master’s Program in Social Data Science

This article is an exploration of the University of Copenhagen and its master’s course in Social Data Science as well as a quick consideration of how collaboration could be facilitated towards efforts in AI Safety.

Social Data Science is a new exciting area of study. This month there was no Wikipedia page on the topic and indeed very few articles on Medium (only one specifically mentioning the topic). I wrote an article on this topic called Towards Social Data Science. There are a few institutions that have begun educating students on master’s level in Social Data Science including (Oxford, LSE and University of Copenhagen). For this article I will focus on the University of Copenhagen considering the possibility of contributing to the field of AI Safety with the combined academic staff present at the University of Copenhagen focused on security issues together with the students attending the master’s course in Social Data Science.

University of Copenhagen

The second oldest university and research institution in Scandinavia is ranked as the best in Scandinavia and the 30th best in the world according to the academic ranking of world universities. It has six faculties one of which is the Faculty of Social Science, which I will focus my attention on. Within these there is a further division into five departments, where research and teaching are carried out in the fields of economics, political science, international politics, management, anthropology, psychology and sociology. Recently however as you may spot there is a focus on Social Data Science.

Social Data Science at University of Copenhagen

If you click onto the tab for Social Data Science you will find further information on the topic. The Faculty of Social Science at The University of Copenhagen defines social data science as the following:

“Social data science is a new discipline combining the social sciences and computer science in which the analysis of big data is linked to social scientific theory and analysis.”

Researchers Affiliated with Social Data Science

Clicking further into the tab ‘our researchers’ you will find that the researchers in this area has a variety of different backgrounds ranging from in order of appearance: political science, anthropology, economics, sociology and psychology. It does not however seem all are actively involved, yet it is fascinating to see this combination of areas coming together listed in this centre.

The research topics listed underneath the different researchers, and I will only mention a few, range from: (1) diplomacy on social media; (2) digital anthropology; (3) data ethics; (4) social data in education; (5) data protection; (6) cyryptomarkets and drug dealing on the Internet; (7) behavioural campaigns; (8) social big data; (9) machine learning and public services; (10) corporate security; (11) military drones; (12) use of AI in military situations; (13) transformative technologies.

I was curious to know the person heading up the program so I inquired further.

The Head of Social Data Science (SODA)

The current head of SODA is David Dreyer Lassen, and he is also the founding director of the Center for Social Data Science (SODAS) as well as co-deputy director of the Center for Economic Behavior and Inequality (CEBI). He is a professor of Economics. His current research is on fiscal policy and budget negotiations. He seems to have been a visiting student and researchers at the University of Harvard while otherwise having a background from the University of Copenhagen.

In 2016 he won an interdisciplinary grant for the project social fabric. The project received funding of 16 million Danish (approx. $2,3 million by today’s value). It additionally involved several other faculties. I ventured to the project site for Social Fabric at the University of Copenhagen.

It seem they collected harvested and analysed data activity from up to 1,000 students’ smartphones as part of the Social Fabric project and it ran from 2013–2017. The departments represented was Anthropology, Economics, Media, Physics, Psychology, Public Health and Sociology. He has a GitHub page where quite a bit of his research and projects are displayed in a neat manner and his CV is extensive.

Research Projects and Groups

The current projects that I could find listed on their website (sodas.ku.dk) are as follows. I have extracted the name and excerpts from the different pages to help you understand in short what all these projects are focused on. I may of course loose important elements in this pursuit, but perhaps it is helpful:

1. Critical Algorithm Lab (CALL). New methods in the meeting between qualitative and quantitative social data. CALL studies researchers who work in a world of websites, social media and large online databases, and identifies the role that social big data can play in new forms of interdisciplinary social science. Headed up by Morten Axel Pedersen and Associate Professor of Sociology Anders Blok.
2. Diplomatic face-work DIPLOFACE. 165 of the world’s heads of state have personal Twitter accounts and two-thirds of them write their own tweets — even during important international negotiations on, say, peace agreements or the distribution of refugees in the EU. They do this because today the public is one of the tools of diplomacy. The project is led by Professor of Political Science Rebecca Adler-Nissen.
3. Mass politics and social Media. Interested in how political opinions are formed. In the Mass Politics and Social Media project, the researchers have “harvested” political discussions from a total of 2.5 million Danish Facebook accounts in the period leading up to the parliamentary election of June 2015 and analysed their tone and content.
4. Microdynamics of influence in social systems. The aim of this project is to gain an understanding of how information is spread and influences people on social networks. Viral processes on social networks influence our opinions, what we buy, and which politicians we vote for, and companies such as Facebook and Google use complex algorithms to ‘nudge’ us to follow their recommendations. The project is headed by associate professor Sune Lehmann Jørgensen, The Technical University of Denmark
5. Social Fabric (2013–2017). For three years, researchers from the University of Copenhagen and the Technical University of Denmark harvested and analysed data activity from up to 1,000 students’ smartphones as part of the Social Fabric project, and the researchers now have precise knowledge of how, when and how much the young people communicate with their fellow students. The main coordinator of the project was Professor David Dreyer Lassen
6. Digital Disinformation. This project provides novel insights into what makes digital disinformation successful in propagating into news media, how ordinary citizens, professional trolls and nonhumans (bots) are implicated, and how receptive different countries and media platforms are to disinformation. Provides new insight into amongst other things information wars. Rebecca Adler-Nissen, Professor in the Department of Political Science, University of Copenhagen is director of Digital Disinformation.
7. Data governance after GDPR. With the acceleration towards a data-driven market economy and public sector, private companies and public institutions are changing their organizational setups to incorporate mechanisms for managing and governing data. The project is led by Kristoffer Albris, Assistant Professor of Anthropology, University of Copenhagen.
8. Connect. Explores how technologies come to shape care, relationships and ‘repertoires’ of being human. The project is carried out in collaboration with the Danish Alzheimer Society, Copenhagen municipality and Åbenrå municipality, and is funded by the Velux Foundation. Led by Nete Schwennesen.
9. Smart Cities Accelerator. In order to optimize the use of renewable energy and reduce CO2 emissions, the project strives to apply context-driven understandings of urban development and human-technology interactions in new solutions within the municipal energy supply systems. Simon Westergaard Lex is co-PI in the interdisciplinary project Smart Cities Accelerator.
10. When the physiotherapist goes digital. Nete Scwennesen studies a sensor based technology, for remote monitored physical rehabilitation. She explores the kind of relationships health care professionals and patients establish with and through this technology and how authority and affect is produced and negotiated in those encounters.

AI Safety, Securitization and Ole Wæver

Ole Wæver is Professor, Department of Political Science at the University of Copenhagen. It seems he is currently studying military drones and transformative technologies. Together with Barry Buzan is famous within theories of International Relations (IR)relating to securitization to the point of this being associated with the Copenhagen school of security studies or simply Copenhagen School in IR. The primary book of the Copenhagen School is Security: A New Framework for Analysis, written by Buzan, Wæver and de Wilde. Ole Wæver is now also focused on climate change as security issue.

In one way due to his focus on military drones and transformative technologies it would make sense to touch upon the growing interest in this area. Particularly now with the growing campaign to stop killer robots. It is not unlikely that this has been discussed already, and he is affiliated with SODA. Considering that artificial intelligence and more specifically the notion of artificial general intelligence has become a growing topic in international relations this does not seem unlikely.

The Copenhagen school is already famous in IR, so it would be rather interesting to see its involvement with SODA. On the surface it seems quite a few of the research projects in connection with Copenhagen Center for Social Data Science (SODAS) could be relevant in this regard perhaps CALL in particular.

My suggestion is that SODAS considers the possibility of formalising a specific research group on AI Safety.

The Degree Programme MSc in Social Data Science

I should move from pure speculation or reflection into specifying the proposed teaching at the University of Copenhagen. The current form on the website of the Faculty of Social Sciences is in Danish. Luckily for you dear reader I am Norwegian, so translating from Danish should be a breeze (albeit with some possible minor mistakes). You could of course Google Translate the page too. This is what they are looking for and competencies they propose to develop:

The modules and layout proposed has the following structure:

I have shortened a few of the descriptions within these different modules all emphasis whether italics or bold is added by me:

The First Half of the First Semester

Starts with an intensive ‘ Social Data Science Boot Camp’. The first two weeks set the stage for the overall educational program.

• Collaboration in teams, and a student culture is founded on the basis of curiosity, community and responsibility.
• Real-life cases for value creation with data. Students work with different types of social data that are put into use through hands-on solutions.
• Mini-field work and gain the first experience with qualitative quantitative approaches.

After that, the course continues with a five-week intensive code course where students learn to use the main programming languages ​​(eg Python) to analyse big social data; Here, too, students are trained to apply these methods to specific issues through teamwork.

First half of the semester, the students also follow a series of lectures, where they are introduced to some of the basic and central issues in social data science and meet their future teachers and supervisors in the education. The course is reserved for students who are admitted to Social Data Science.

The Second half of the First Semester

consists of the two courses: ‘ Social Data Science’ and ‘Data Governance: Ethics, Law and Politics’ (see Figure 2 above).

1. In ‘ Social Data Science’ , students work with quantitative data in challenging digital formats and analyze and visualize human behavior and interactions by combining technical and social-analytical skills.
2. These newly acquired competencies are then actively involved in the course ‘Data Governance: Ethics, Law and Politics’, with both courses being replaced in the form of an integrated examination paper, combining ethical, legal and political principles and perspectives with more technical and quantitative skills to identify, manage and analyze large and complex social data sets in relevant organizational contexts.

The second semester

follows two courses in ‘Advanced Social Data Science’ .

Part 1 focuses specifically on methods to investigate behaviour, networking and ideas and is integrated into the exam with the course ‘Re-tooling Social Analysis: Behaviours, Networks, Ideas’ . The latter aims to provide students with theoretical and reflexive skills to understand, translate and deal with analytical challenges and ethical implications by working with various social data, including relating to the validity of social data and analyses thereof.

Part 2 deals with unstructured data, including text and images, new data forms and advanced data structures. The course is integrated with ‘Digital Methods, Ethnography and Content Analysis’, where students learn in practice how to use a wide range of qualitative methodologies and to integrate them with quantitative approaches in order to perform qualitative quantitative analyses of social science issues.

The integrated exam is a self-selected project that brings these different methods and analytical approaches together in the same study, whereby this exam serves as a model for what students can choose to do in their later thesis.

Furthermore, the form of the examination is ‘authentic’ in the sense that the formats in the integrated exams mimic the product deliveries that future candidates are expected to deliver in their future workplaces (see below in the fourth semester).

The third semester of the program

Consists of a compulsory project-based seminar, where the students independently choose to work on a subject they wish to specialize further in the framework of social data science.

There will be workshops and learning activities linked to the course, which will be prepared in collaboration with external partners. In addition to the project seminar, students can choose between completing fieldwork, a project-oriented course or following electives offered by other programs.

There will be designated packages of electives from the University’s other faculties, which the students can orient themselves to if they have specific academic interests in each of the faculties’ areas. Students who wish to undertake fieldwork, complete a project-oriented course or exchange may obtain exemption from the compulsory seminar.

The Fourth Semester

The program concludes with a thesis in the which will combine at least two of the four academic components of the program.

The thesis can be replaced in the form of a traditional thesis or, for example, as a product delivery.

A product delivery can originate from a collaboration with an external partner and reflect its specific need for the development of a specific product that the specialist student prepares. Possible product deliveries may include:

• A commented and analyzed dataset with documentation and validation
• A commented and documented algorithm
• Combinations of a data ethical protocol, data and implementation strategies, needs studies and user studies
• Validation of a method of data collection and / or analysis
• Empirical testing / testing of theory / model on new data forms
• Development and conduct of experiment / A / B testing
• Predictive models, including business intelligence and artificial intelligence

This ends with an oral defence.

Needs for This New Education

They made a study of the labour market prior to the creation of this course.

“The results of the study show that not only is there a great need for graduates with a Social Data Science competency profile, but that significant demand is increasing. The results are based on analysis of 1,600,000 job listings for the period 2010–2018 at Jobindex, Denmark’s largest and most comprehensive job bank with vacancies in both private and public sector. The data base represents all vacancies posted on the Jobindex during the period.”

Here are some key results from the study that the University of Copenhagen undertook:

1. Since 2010, the need for Social Data Science graduates has doubled in the labor market.
2. In 2010, 2% of social science job vacancies demanded what we identify as Social Data Science competencies, while demand in 2018 has risen to 17%, ie a 15 percentage point increase.
3. In 2018, 50 posts will be posted per month where employers are looking for employees with skills that integrate classic social science and data science. In 2010, 4 positions were posted a month, demanding this competency profile.
4. Both the private (65%) and public (35%) sectors need these skills.
5. The demand for Social Data Science graduates is almost twice that — and is growing significantly faster than — the demand for commercial data science competencies.

“The strength of candidates with a combined classical social science and data science profile is, among other things, according to the customers, that managing large data sets is not an end in itself, but that the understanding of data — including for the opportunities they have for knowledge and analysis as well as for the societal and / or business consequences of the knowledge and analysis — is included and used in the work itself to structure and manage data.”

This analysis has led to the identification of three competencies, which the customer comments generally refer to, which generally refer to the upcoming Social Data Science candidates as:

• ‘Translators’
• ‘Analytical translators’
• ‘Analytical design’
• ‘Analytic leadership’

AI Safety Needs Social Scientists

In a paper called AI Safety needs Social Scientists Geoffrey Irving and Amanda Askell is arguing that long-term AI safety research needs social scientists to ensure AI alignment algorithms succeed when actual humans are involved. I wrote an article previously called Social Scientists and AI focused on this paper, however this excerpt from their abstract is striking.

Properly aligning advanced AI systems with human values requires resolving many uncertainties related to the psychology of human rationality, emotion, and biases. The aim of this paper is to spark further collaboration between machine learning and social science researchers, and we plan to hire social scientists to work on this full time at OpenAI.

In the development of artificial general intelligence I believe that this is vitally important, of course since as I study in the social sciences with computer science, however also as a concerned citizen.

Issues arising must be handled responsibly and part of this requires challenging interdisciplinary work between the social sciences and the natural sciences. Much like ‘engineering’ and ‘science’ these labels can be called social constructs, yet they come with a degree of praxis connected and such can be important signifiers for collaborative frameworks.

Conclusion

The Social Data Science MSc at the University of Copenhagen does seem rather interesting to look into. Particularly if you are heading towards AI Safety due to its close ties with an established environment of researchers and research groups connected to security such as Ole Wæver with his International Relations theories of securitization. My suggestion to Copenhagen Center for Social Data Science is that they consider the possibility of creating a formalised group for AI Safety.

There are not many universities yet that offer this specialised form of education, so if you are as interested in this specific area you could consider checking it out either as a student or to keep an eye on the curriculum and teaching methods. If you run a company looking for these skills it may also be well worth to keep an eye out for the graduates or research resulting from this investment.

Thank you for reading. This is day 58 of #500daysofAI. I write one new article every day on the topic of artificial intelligence.