Ethics and Explainability for Responsible Data Science (EE-RDS)

Ethics and Explainability for Responsible Data Science (EE-RDS)


Publishing Date: 4/15/2021 1:00 PM

art Conference hosted by:

  • Data Science Across Disciplines (DSAD) Research Group, Institute for the Future of Knowledge, University of Johannesburg
  • Perception Robotics and Intelligent Machines Research Group (PRIME)University of Moncton
  • Sponsorship is being provided by the National Integrated Cyberinfrastructure System (NICIS)

Dates: 27 – 28 October 2021 

Registration: Please register for the conference here. All participants as well as presenters must register. If you are planning to submit an abstract, then you will need to do so as per the instructions below.

art

KEYNOTE SPEAKERS

Yoshua Bengio (Department of Computer Science and Operational Research, Université de Montréal, IVADO, CIFAR, Scientific Director - Mila)

Geralyn Miller (Sr. Director of the AI for Good Research Lab at Microsoft) 

Mark Parsons (Editor in Chief, Data Science Journal, University of Alabama)

Lianglin Hu (Director of National Basic Science Data Center, Deputy Director of the Big Data Department of CNIC)

art
THEMES

Is Data Science a new approach to solving problems, one that applies across disciplines as various as physics, sociology and linguistics? Or are machine learning, deep convoluted neural nets, and other exciting phrases just statistics on steroids?

Recent developments in Data Science broadly construed, and the products these have yielded (or promise to yield) are undeniably exciting: identifying and predicting disease, personalised healthcare recommendations, automating digital ad placement, predicting incarceration rates, and countless other tools have attracted a lot of attention. But what about the process behind these products? Are these amazing feats based on traditional scientific discoveries? Or does the problem-solving approach which is being implemented have an even wider range of applicability than we could imagine? While the Sciences and Engineering are driving the field, traditional Humanities and the Social Sciences are also experimenting and contributing to a growing body of knowledge around the use of data. This conference seeks to understand the nature and significance of data science for traditional modes of inquiry across the full spectrum. We also seek to interrogate underlying ethical issues that arise not only in research but also when data science is relied on in decision-making – this is where notions of explainability, fairness and discrimination form part of the practical application of responsible data science.

As a launching event of the Data Science Across Disciplines Research Group at the University of Johannesburg, this conference brings together reflections on both the actual and potential impact of data science across disciplines and sectors. Submissions are welcome from any disciplinary background, with a focus on scientific contributions, conceptual themes, and reflections within the areas of:

  1. Responsible Data Science: Reliable and Trustworthy approaches for data engineering, data science and modern machine learning.
  2. Algorithmic Fairness, Transparency, and Explainability.
  3. Social and Ethical aspects of Responsible Data Science.
  4. Use cases illustrating the cross-disciplinary nature of the field of Data Science.

All papers must be pitched in a suitably accessible way and speak to the cross-disciplinary nature of the event.

art

ABSTRACT SUBMISSION

Please ensure that you have registered for the conference here before submitting an abstract. 

Abstract Submission: Please submit your extended abstract on the Microsoft CMT website and ensure that you use the IEEE abstract template provided here: IEEE Template.

Note: You may use either the LaTeX or Word template but your extended abstract must be a minimum of 4 pages long and in .pdf format
When you submit your extended abstract you will be asked to indicate whether you would be interested in publishing your work in IEEEXplore proceedings at a minimal fee. The authors of submissions of suitable quality will be contacted at a later stage should they indicate an interest in doing so.

Abstract Due Date: 30 September 2021

art
SCIENTIFIC PROGRAMME

The Scientific Programme does not only encompass presentations by Keynote Speakers and registered participants, but also a Problem-Solving panel during which experts in the field will discuss and debate a possible solution to a stipulated problem within the field of Data Science. All conference participants and especially Postgraduate Students are encouraged to attend this panel debate and will be allowed to engage via the chat function.

27 October 2021
​13:00 – 13:10
​Opening Speaker
​13:10 – 13:30
​Speaker
​13:35 – 13:55
​Speaker
​14:00 – 14:20
​Speaker
​14:25 – 14:45
​Speaker
14:50 – 15:50
​Panel Discussion
​"Trustworthy AI and Lending"
Chaired by Kush R. Varshney
​15:50  16:00
Break
16:00  17:00
​Plenary Speaker: Yoshua Bengio
​"AI for Social Good"
​17:00  17:20
​Speaker

​17:25  17:45
​Speaker
​17:50  18:10
​Break
​18:10  18:30
​Speaker
​18:35  18:55
​Speaker
19:00 – 20:00
​Plenary Speaker: Mark Parsons

28 October 2021

​13:00  13:10
​Opening Speaker
​13:10  13:30
​Speaker
​13:35  13:55
​Speaker
​14:00  14:20
​Speaker
​14:25  14:45
​Speaker
14:50  15:50
​Panel Discussion
​"Data Stewardship and Responsible Data Science"
Chaired by Louise Bezuidenhout
​15:50  16:00
​Break

16:00  17:00
​Plenary Speaker: Geralyn Miller
​17:00  17:20
​Speaker
​17:25  17:45
​Speaker
17:50  18:10
​Break
​18:10  18:30
​Speaker
​18:35  18:55
​Speaker
19:00  20:00
​Plenary Speaker: Lianglin Hu 

PANEL DISCUSSION

As a part of this conference, we will host a Problem-Solving Panel Discussion where a group of specialists will consider a problem of real-world importance; they will clarify the issue at hand, discuss possible issues involved, consider the tools at their disposal and ultimately design and argue for a feasible solution.

On each of the days of the conference, 60 minutes will be set aside for a panel discussion on a particular problem or issue related to the theme(s) of the Conference. Panel members will be assigned by the Scientific Committee of the Conference, and attendees will be allowed to sign up to attend as a part of the audience.

27 October 2021 (14:50 – 15:50) 
Panel Discussion: ​"Trustworthy AI and Lending"

Chair: 
  • Kush R. Varshney, Distinguished Research Staff Member and Manager, IBM Thomas J. Watson Research Centre
Panel members: 
  • Jiahao Chen, CTO at Parity AI
  • Moise Busogi, Carnegie Mellon University (CMU) Africa
Topic: 
In this panel discussion, the discussants will be examining home mortgage lending approval decisions carried out by machine learning systems. They will consider the various phases of the development lifecycle (problem specification, data preparation, modelling, evaluation, deployment) from the perspective of trustworthy AI. They will critically examine the fairness, explainability, robustness, and transparency of a solution that could be created.

28 October 2021 (14:50 – 15:50) 
Panel Discussion: ​"Data Stewardship and Responsible Data Science: Lessons from the CODATA-RDA Schools for Research Data Science"

Chair: 
  • Louise Bezuidenhout, University of Cape Town 
Panel members: 
  • Hugh Shanahan, Royal Holloway University of London
  • Shanmugasundaram Venkataraman, Digital Curation Centre
  • Joy Davidson, Digital Curation Centre
  • Sanjin Muftic, University of Cape Town
  • Naniki Maphakwane, Botswana Open University
Topic: 
Data-driven research relies on a range of expertise within research communities. This has led to the emergence of the data stewards, individuals who provide oversight or data governance support within organizations and ensure the quality and fitness for purpose of data assets including the metadata. The CODATA-RDA Schools for Research Data Science, together with FAIRsFAIR, have developed data stewards training workshops that focus on training on responsible/open research and research data management. This panel will discuss this training in detail, and include a broader discussion about the benefits and challenges of educating data stewards and their impact on responsible research.

Appropriate links are provided here:
CODATA-RDA Schools for Research Data Science: Link
Example of upcoming data steward school: Link
Data Steward Schools collaboration with FAIRsFAIR: Link

POSTGRADUATE CHALLENGE

PG Students from all educational backgrounds are invited to take part in the EE-RDS Conference Challenge!  This year's challenge is the effective detection of COVID-19 via Chest X-ray images. This is an opportunity for aspiring data scientists to work on a problem of global importance and compete for the chance to win one of two laptops sponsored by the National Integrated Cyberinfrastructure System (NICIS). We await your innovative ideas and skilful programming!

Details to the challenge may be obtained here: https://cxr-covid19.grand-challenge.org/
Participants  need to join to get access to the data. The train/validation data set is ready and the test data set with a submission sample will be posted soon.

ORGANISING COMMITTEE

Chair of Committee: 

  • Charis Harley, Head of DSAD, Faculty of Engineering and the Built Environment, University of Johannesburg

Committee:

  • Moulay Akhloufi, Head of the Perception Robotics and Intelligent Machines Group, Department of Computer Science, University of Moncton
  • Terence van Zyl, Institute of Intelligent Systems, University of Johannesburg
  • Anwar Vahed, Director at NICIS Data Intensive Research Initiative of South Africa