This is exactly why I changed careers to #DataScience. I wanted my work to have a positive impact and I’m so glad to be accomplishing this at #PredictWise 👇🏽👩🏽💻
2/4 Break grounds in combining attitudinal/survey data with behavioral/telemetry data to build stronger, more accurate, more dynamic Machine Learning models of political attitudes
4/4 Finding new ways in validating and re-validating our Machine Learning models in the field in rigorous randomized control trials. And, above all, pushing for always-on data pipelines, now that 2020 is (firmly) behind us more important than ever
3/4 Re-think identity resolution, and the related precision-recall trade-off, to build high-precision data bases of unregistered Americans and using this data to power text banks with accurate cell-phone numbers, and mobile-first advertising campaigns
1/4 In the 2020 cycle, @PredictWise was at the forefront of technological innovation, bringing sophisticated data to bear on behalf of progressives up/down ballots. @NikhGarg has written up all that he and @gioamorim accomplished last year: https://t.co/JGgJtmMZND
4/4 Our targeting leads to measurable results (far beyond the typical “we worked with campaign A, and campaign A won”). Together with TruthNotLies, we moved a Trump 2016 universe away from Trump (stat. significant; RCT). Thank you for all @PredictWise to making this possible!
2/4 Better targeting: our scores are based on always on survey AND behavioral data collection, dynamic Bayesian Machine Learning built on GPUs, and projection/extrapolation built on CuBLAS, allowing us to score all Americans within hours
3/4 Platform independent technology means that our targeting fed right into demand-side platforms, not just Facebook, providing a crucial alternative when Facebook shut down political ads, and performing better
1/4 In the 2020 cycle, @PredictWise was active at the national level, and in all major states. We helped progressives w/
1. Better targeting through machine learning and always-on data collection
2. Platform-independent technology
https://t.co/vaIYxzhy4n
4 suggestions Democrats by @KonitzerTobias (1) fund spending off-cycle, rather than all at end of election campaign (2) move digital beyond Facebook for stability & reach (3) understand why voters support causes & parties (4) address them directly & continuously all cycle long
While these have been long 36+ hours with lots uncertainty left, I believe there are lessons from the election -- 4 pathologies Dems would do well to address. Thread 1/5 - full piece @PredictWise https://t.co/ig2DD5dGDI
The whole team @PredictWise has paved the way for a new era of digital outreach, focused on thinking beyond Facebook. Hard to overstate the impact of this work, especially over the last few weeks of the cycle. Thank you!
Wrapped up my last delivery for the cycle earlier this morning. Can't overstate how much working on the election @PredictWise have kept me (somewhat) sane these past 7 months! More soon on the cool and impactful stuff we did.
Hat-tip to @KonitzerTobias@PredictWise who are tracking 250+ questions each week, all geared towards underlying issues & values, to better understand true nature of electorate, rather than fleeting position on short-term policy debates.
4 biggest policy issues of 2020: healthcare/COVID, economy, racial justice, climate change. Lots of polls try to capture snapshots of sentiment on these policies. @PredictWise has been tracking underlying sentiment for on these topics for years. Here are a few thoughts:
new research @PredictWise : estimating size of news audiences, or the number of viewers for Trump's covid briefing, with surveys just does not work -- behavioral data is needed: https://t.co/ptMJ1HZnKU
Proud to officially launch SIGNAL, our digital audience builder @PredictWise, allowing you to target at intersection of attitudes & behaviors (real time location? check! Reaching COVID-decree adhering Republicans? check!) Read more here https://t.co/xYzcnY2d2z
.@PredictWise has run a battery on risk & investing for a few years, and I am annoyed that this concept of "We all Benefit" from the stock market going up has not shifted in the face of reality: yet.