Subtle Cues

Substratum tracking and listening engine for understanding subtle cues to predict propagation of infectious diseases 

Cordula A. Robinson, Ph.D.. Senior research scientist at Northeastern University’s Kostas Research Institute, leading geospatial AI knowledge prototype development

The world is currently experiencing the devastating impacts of the COVID-19 pandemic as a health crisis, an economic crisis, a humanitarian crisis, a security crisis, and a human rights crisis (UN). COVID-19 is not vanquished by the immune system and everyone is assumed to be susceptible. The COVID-19 pandemic has led to a dramatic loss of human life worldwide (close to 2.8 million deaths) and presents an unprecedented challenge to public health, food systems and the world of work (WHO) that we never want to experience again. 

Primary customers include government, emergency and/or public health authorities. Current investigation concerns fine-grained geographical variation and substantive insights that address predictive accuracy with a generalizable toolchain. Present focus is the US. Through an open BAA (Broad Agency Announcements), Northeastern/KRI was awarded a 1-year, $1million contract to lead Natural Language Processing (NPI)-related research, constrained by a common operating picture, to overcome the difficult challenges presented by the COVID-19 pandemic. Long-term goal is to develop an application that can be used by the general public to retrieve accurate insights into potential disease outbreaks for their own cognizance.

Governments, emergency officials, public health authorities, even concerned in-place and traveling individuals, seek accurate insights into potential infectious disease outbreaks for proactive response and mitigation. Subtle Cues can help by listening to communities to offer early-warning perspectives of local virus manifestations (potentially pre-medical diagnosis). The beauty of the Subtle Cues is that it listens indiscriminately - we don’t care who you are, only where you are, and what you are saying. Our goal is to become an enabling tool as part of a software-enabled fight to mitigate future catastrophes of the magnitude the world is presently experiencing with COVID-19, where impacts are experienced not only as a health crisis, rather also as an economic crisis, a humanitarian crisis, a security crisis, and a human rights crisis (UN). 

To date, the world has experienced approximately 2.8 million COVID-19 related deaths, and 130 million cases of COVID-19 infections (CDC). Older adults, and people with severe underlying medical conditions, are at highest risk for developing serious complications and long-term health problems (CDC).  World economies are also struggling (International Monetary Fund). Implementation of data-driven analytics can help minimize such calamitous repercussions when incorporated as a response agent by relevant agencies, industries, and/or individuals. Subtle Cues is a new analytical technology that empowers responses to infectious disease risk through case identification constrained by fine-grained geographical variation and insight into opinion and behavior. Subtle Cues emphasizes substantive insights occurring in a neighborhood, therefore, moves beyond the realm of pure predictive accuracy and into the domain of behavioral response. It is designed to be a collaborative data-driven system that can support real-time virus outbreak predictions, potentially before medical diagnosis.

Subtle Cues is a substratum listening engine and proof-of-concept technology that, can help mitigate future pandemics as an enabling tool. Current focus is stabilizing the technology (while finding early adopters interested in what we were doing) and considering a larger future for financial sustainability. Present focus is Massachusetts, USA, where a plethora of data are available to help understand the propagation of the COVID-19 virus as it impacts 7 million people. The granular focus enables high-quality information to be retrieved for diverse populations, thus fuel an effective AI solution. From this baseline, demonstration and validation in new environments and new languages will be sought, e.g., Brazil, also India and China (countries from which team members hail). While relying on translation for multi-lingual approaches is viable, native wisdom and experience are a key ingredient to capture culture and nuance. Our model is to interface Subtle Cues for highly customized offerings needed by governmental emergency and public health authorities worldwide. More generic, ubiquitous offerings for everyday citizens through smartphone applications would allow for horizontal scaling - global mobile apps generate > $31.9 billion  in revenue via paid downloads and in-app advertising (Statistica).  

Subtle Cues is designed to be a multi-lingual substratum tracking and listening engine that detects signals to predict the propagation of infectious diseases and virus manifestation as they unfold locally across the world. By “listening” to social media discourse, anonymized and non-discriminatory insights can be gleaned at a community level.

Insights reflect public deliberation, behavioral responses, and pre-infection risk behaviors that correlate with virus leads and lags. Assessing and evaluating words, phrases, and speech behaviors is used to anticipate threats of virus spread and help predict the propagation and impact of infectious diseases. Textual correlates that fall within COVID-19’s 2-week incubation period provide early-warning for virus outbreak.

The Subtle Cues engine is designed to accommodate greater sophistication and collaborate with other data-driven analytic systems to become part of the software-enabled fight for ground-breaking solutions for future health emergencies. Our goal is to mitigate the devastating spread of pandemics such as those currently experienced with COVID-19, which presents an unprecedented challenge to public health, food systems and the world of work (WHO).

Barrier 1 - Seedling financial support. We are a recipient of a $1million COVID-19 NPI seedling grant to develop integrative research methods with US focus. Barrier 1 is traversed.

Barrier 2 - Constitutes site and mechanism for expansion. I have filed disclosures with Northeastern University for provisional patent applications and a long-range view of productizing the solution to begin a small-business venture. The Kostas Research Institute at Northeastern University, the home of this work, sits on the Innovation Campus at Burlington MA (ICBM). ICBM is home to 17 early stage companies with room for more. Barrier 2 is traversed.

Barrier 3 - Growth and development: Fund raise and establish capitol for expanded team to develop AI knowledge prototype for next gen AI toward productization and revenue generation. Technical: Remain ahead of the curve to advance AI technology and keep improving through iteration. Cultural/cyber: mainstream acceptance of software as a tool to enhance individual value of life without misappropriation. These barriers are continuous.

Kostas Research Institute at Northeastern University and Northeastern University: home institute

Fiocruz, Brazil: data integration for early-warning identification and characterization of respiratory-transmitted viral outbreaks with pandemic potential

 RAMP, Cambridge UK: potential opportunity to create an add-on behavioral model extension

1. Growth and development: Fund raise to introduce multilingual capability to Subtle Cues to validate and expand capabilities into other countries, for other infectious diseases.
2. To establish capitol for expanded team to develop our AI knowledge prototype toward productization and revenue generation. This might be through the creation of our own small business, at Northeastern's Innovation Campus in Burlington MA, and/or through partnerships with public health or risk-assessment software companies interested to add innovative behavioral tech to their software arsenal.
3. Data privileges for model augmentation, e.g., international data from South America, Europe or Asia, particularly as it pertains to COVID-19 case counts at a town level; localized daily aggregated Facebook textual data, or Google-trend data, at the town or smaller level (though the API). 
4. Professional networks/ customer base/awareness

Advisor /Expertise/Ambassador/Funder: Fiocruz, Tencent, Aviva, Infosys - in view of expanding and developing Subtle Cues for other countries and languages, and to other viruses.

Privileged Data Access/Funder/Expertise: Google/Facebook - for access to day-town data for specific geographies. Subtle Cues benefits from additional data sources, such as aggregated Facebook textual data and Google search data where text is sufficiently dense to fit models at the day-town level.

Advisor/Expertise/Ambassador/Funder: BlueDot – Subtle Cues would welcome continued conversations with BlueDot since this style of tech is currently not part of their software suite.

Advisor/Expertise/Funder: Global Virome Project, Doctor Evidence, Aviva, Optum, Palantir - for more advanced options in view of productization with a public health / insurance organization. 

Funder: Clinton Health Access Initiative, Bill and Melinda Gates foundation.

Privileged Use of Tools for expansion: Microsoft.