Bridging the divide between social and technical systems: techniques and current research

Abstract

Leeds Critical Data Studies Group (LCDSG) is hosting a joint seminar with AimTech, the Socio-Technical Centre (STC), Corporate Social Responsibility Group (CSR) and the Centre for Decision Research (CDR) aimed at bringing together researchers from across the University of Leeds who have an interest in the way that people, analytics, data and the accompanying technologies interact.  We have interesting talks on the use of Practice Theory, including the practical application of Activity Theory and The Mangle of Practice (both have led to publication for the academics and practical solutions/improvement for the practitioners).  We also have a talk from LCDSG lead, George Ellison on “The Professionalisation of Data Science: risks, benefits and opportunities".

Please register to attend at the Leeds University Business School events page - Bridging the divide between social and technical systems: techniques and current research

Using Activity Theory to understand Police use of Social Media

Emma Forsgren, Emma Gritt and David Allen (AIMTech)

Social media has transitioned from leisure-based communication into a platform for professional work across many organisations. The police force is one such organisation that is integrating social media into its practice, especially as a means for community engagement. This effort creates new opportunities to engage with the public, whilst also generating challenges that conflict with the organisational culture of the police, and leads to complicated situations relating to police activities. While many studies rely on retrospectively analysing tweets to understand police use of social media, we suggest that more research should take a qualitative approach and explore the socio-cultural aspects that surround police use of social media. This will enable us to better understand how police adopt and use social media and how this fits into their work practices. This work aims to understand the influence of social media on police activities through the lens of activity theory. As part of a larger study of the U.K. police force, this work used a multi-method approach to gather qualitative data from a variety of roles and positions. Key findings highlighted two contradictions that had an influence upon the activity of community engagement, in terms of (1) rule-based organisation vs. amorphous social media, and (2) limited resources vs. maintenance and engagement in evolving platforms.

The Professionalisation of Data Science: risks, benefits and opportunities

George Ellison PhD DSc (LIDA Deputy Director - E&T)

'Data Science' remains contentious in every way possible - what it is, what it does, who can do it and whether it should be regulated. This brief overview of ongoing work examining contemporary attempts to 'professionalise' Data Science will explore the risks, benefits and opportunities this emerging new community of practice might face; along with what such attempts might tell us about the changing nature of a 'professional' in a globalising neoliberal market place. I will argue that Data Science introduces (at least) two novel risks for practitioners and society related to: the ownership of publicly and/or commercially accessible data; and the opacity of machine learning algorithms used to inform decisions for good or ill. These create an imperative for Data Scientists to embrace professionalism in some form if they are to retain the independence required for creative innovation. What form such 'professionalisation' might take will be discussed with reference to traditional, contemporary and informal, distributed forms of professional affiliation and (self-)regulation.

Activity Theory as a method to study technology-mediated complex collaborative work.

Aleksandra Irnazarow, PhD Candidate and Research Fellow in Data Analytics, University of Leeds, UK

This talk will explore applications of practice-based theories, in particular activity theory, to diverse contexts within organisational environments, including the design and development of complex engineering systems. It is based on a PhD research project involving case studies within three high-value engineering organisations. The talk will discuss the use of activity theory to examine relationships between complex collaborative work practices and information technologies through focusing on questions relating to how teams of professionals make decisions and the role that technology plays in their decision making.

A Mangle IN Practice as a framework for studying Action Research, Design Science and Participative Modelling

Christina Phillips PhD, Business Analytics/Statistics Lecturer, Leeds University Business School.

This talk follows from work on a 4.5 yr longitudinal case study in a complex pharmaceutical manufacturer with particular focus on operations and supply chain.  The project encompassed many techniques and demanded a multidisciplinary approach and application of modelling and analytics.  Studying this complex web of socio-material interactions in a real time environment needed lenses which have broad boundaries yet provide enough structure for recoverable information to emerge.  We will use this case to discuss the Mangle of Practice adapted as a Mangle IN Practice for use in researching and observing technical interventions in real world settings.

Artificial Intelligence for Language,

Professor Eric Atwell, AI research group, School of Computing, University of Leeds

Data analytics and corpus linguistics can classify and “understand” text snippets, eg

• Quran and Bible verses: find similar/related verse pairs, key themes or topics;
• classify Twitter tweets and Facebook posts by dialect, sentiment, …
• patient records, eg predicting time and cause of death from verbal autopsies
• detecting terrorist activities by analyzing text from “subjects of interest”
• sentiment analysis of customer reviews: good v bad products

Ethical issues surrounding text analytics and corpus linguistics? De facto: NONE.

We collate text snippets in a CORPUS; if an “owner” objects, remove their snippet.

How can text analytics help build algorithm familiarity?

Text data, eg tweets, seem more “natural” and “understandable” than numeric data.

Text analytics results, eg sentiment classification, seem easy to understand.

Text analytics research lets us study “messy social spaces” eg analysis of tweets, FaceBook posts; methods and results are publishable in best journals, eg:

• Language Resources and Evaluation Journal (Springer)
• International Journal of Corpus Linguistics (John Benjamins)
• Natural Language Engineering Journal (Cambridge University Press)