My primary research and professional objective is to use quantitative methods to study the impact of technology on human interactions, specifically political discourse and civic engagement.
In our projects, I apply statistical techniques, machine learning tools and NLP methods to study discourses, behaviors and biases in digital communities at scale. More specifically, I aim to develop software tools for tasks such as gauging political polarization, studying hate speech, misogyny and racial and religious bias online, and social media election campiagn techniques in India and the US.
Extreme Speech in Indian politics. Using the affordances theory, we analysed the differing consequneces of using 'extreme' speech for politicians and regualr citizens in India. We annotatted 469 hashtags used by a random sample of 997 Indian politicians as extreme and controversial, based on definitions from prior literature. We find that use of extreme hashtags gave a gain of upto 23 percent in retweets earned by politicians and led to no legal action. We juxtapose this with the prosecution and extra-judicial detention of common citizens that use similar speech against establishment figures using colonial era free-speech laws. This work was in collaboration with Noopur Raval , Han Zhang (UMich), Sunandan Chakraborty (IUPUI) and Mugdha Mohapatra.
Party/Leader centrality We qunatify the centrality of leaders and parties in the 2019 Indian general election campaign by studying their Twitter and Facebook profiles. I am working with Ashwin Rajaesingan at the University of Michigan's School of Information.
Thematic groupings in election propaganda - We are studying the thematic preferences deployed in Twitter campaigns by over 7400 politicians of India's two main parties, the BJP and the INC, by analysing the correlation, cooccurence and temporal variance of hashtags in their tweets. We This work is done jointly with Ramaravind Kommiya Mothilal, Monojit Choudhury and Kalika Bali. A graph plotting the partisanship of hashtags can be seen here. The partisnaship metrics are based on work by Hemphil et al.
NivaDuck / PoliTwictionary - At MSR India, we are collecting a database of Twitter accounts of public figures relevant to Indian politics. We have built NivaDuck - Marathi word for selector - a machine learning classification pipeline that can find Twitter handles of politicians in a given country by leveraging the Twitter bio, tweet data, hashtags used and friend network of a seed set. NivaDuck has so far identified in excess of 15000 Indian politicians and 8000+ US politicians at the national, state and local levels. We are working continuosly to improve NivaDuck's performance and functionality, such as identifying party, state and level in the organisational hierarchy.
Twitter in the 2019 Indian General Elections: Trends of Use Across States and Parties. Joyojeet Pal, Anmol Panda.
Economics and Political Weekly. Vol. 54, Issue No. 51, 28 Dec, 2019. [Online].
Chor or Chowkidar: who won the ‘chowkidar’ battle on Twitter during the 2019 Indian elections. Anmol Panda, Joyojeet Pal.
Abstract accepted for oral presentation at Workshop on Bias, Disinformation, Misinformation, and Propaganda in Online News and Social Media at SocInfo 2019, Qatar.
Affording Extremes: Incivility, Social Media and Democracy in the Indian context. Anmol Panda, Han Zhang, Noopur Raval, Sunandan Chakravarty, Mugdha Mohapatra, Joyojeet Pal. Microsoft Research India.
Paper accepted at the Practical Social Informatics: Collaboration Across Fields, Sectors, and Borders workshop at ASIS&T 2019, Melbourne, Australia. New!
Typological analysis of extreme speech related tweets in the context of Indian electoral politics. Anmol Panda, Ramgopal Chandrashekaran, Shubi Agarwal, Joyojeet Pal. Microsoft Research India.
Presented at the International workshop on Global Perspectives on Extreme Speech Online, 10-11 December 2018, Ludwig-Maxmilan University, Munich, Germany.
[ Workshop website ]
The use of social media in the 2019 General Elections. Joyojeet Pal, Azhagu Meena, Drupa Dinnie Charles, Anmol Panda. Published in the Seminar Magazine.
[ Link ]
How #BJP fused with #StrongIndia in #2019. Joyojeet Pal, Anmol Panda, Faisal Lalani. Full page story for the Mint newspaper.
[ Article on livemint.com ]
Narendra Modi matinee show: Inside India's celeb Twitter. Joyojeet Pal,Anmol Panda. Full page story for the Mint newspaper.
[ Article on livemint.com ]
Elections 2019: What tweet for tat tells you about Rahul Gandhi. Joyojeet Pal. Contributions by Andre Gonawela, Anmol Panda, Ramarvind K.M. and Ajai Sreevatsan. Full page story for the Mint newspaper.
[ Article on livemint.com ]
On Twitter, Sachin Pilot is CM and Shivraj trumps Modi in MP. Joyojeet Pal, Anmol Panda, Ramgopal Chandrashekaran Article on the 2018 state assembly elections for Livemint.com.
[ Article on livemint.com ]
COP: Compiler Optimizations to Reduce Memory Stalls for Network Pipelines Written in P4. Shailja Pandey, Ankit Bhardwaj, Anmol Panda and Sorav Bansal. Poster in 15th USENIX Symposium on Networked Systems Design and Implementation (NSDI), April 2018.
[ Poster abstract ]
We are also presenting our poster at the NetApp University day in Bengaluru. [ NetApp poster ]
A Comparative study of GPUVerify and GKLEE. Anmol Panda, Phillip Ruemmer, Neena Goveas 4th International IEEE conference on Parallel and Distributed Grid Computing, JUIT, Waknaghat, HP, India, December 2016.
[ Paper ] [ Undergraduate thesis ] [ Slides ]
I received a scholarship from Erasmus Mundus (now Erasmus+) NAMASTE program to visit Uppsala University in Sweden as an exchange student during the Autumn 2015 semester. At the Department of Information Technology there, I completed four courses: Combinatorial optimisation using constraint programming, Real time systems, Graph Theory and Human Computer Interaction. During one of the assignments of the Real time systems course, we built a path following bot. A video is attached here.
As a TA for MPI, I worked with three more TAs for a tutorial section of 120 students for two hours once every week. We prepared solutions for tutorial questions, aided the students in solving them and resolved the queries. Lastly, we invigilated during the lab exam of the course.
As the TA for EPS, I was present in every lecture and each of the five activities (story telling, impromptu speech, self-intro/prop, debate and persuasive speech) of the course. I prepared topics for each activity, group lists for group activities, assisted the instructor in grading and recorded the scores. Moreover, I presented a sample debate along with eight other students and a sample persuasive speech. Lastly, I mentored students to improve their English proficiency, content structure and delivery. We considered factors like voice modulation, expressions, gestures, pace and pauses, and eye contact.
As the sole TA for one section (ninety students) of TRW, I graded quizzes, assignments and reports. I also set the topics for the students' assignments and reports.
At IIT Delhi, I worked with Ankit Bhardwaj and Shailja Pandey to design an implement compiler optimisations for packet processing applications. These optimizations source their gains from memory level parallelism through batching, efficient scheduling, and cross-producting of lookup tables. Our poster on the same was accepted at NSDI 2018. It has also been submitted as a full paper at OSDI 2018.
The MLP benchmarks and experiments are detailed in this document MLP
During my final semester at BITS Pilani Goa, I worked on a comparative study of two GPU verification tools to assess the programming bugs (data races and divergent barriers) that they reported, benefits and disadvantages of each, and their recommended usage in software development.
A Survey of verification tools for GPU software [Undergraduate thesis]
As a Mitacs Globalink research intern at the University of Northern British Columbia in Canada, I worked under the supervision of Alex Aravind on a project to automatically seed farms. During the twelve week project, along with one more intern, I evaluated four GPS free localisation algorithms for wireless sensor nodes and simulated two in Java
Microprocessors - Programming and Interfacing
Effective Public Speaking