A Human Touch in the Age of AI

In an era dominated by AI research, the need for human insight remains crucial. This was clearly demonstrated in a stunning study led by Dr. Veselin Kostov from the SETI Institute and NASA’s Goddard Space Flight Center. Under his guidance, a remarkable collaboration of nearly 1,800 citizen scientists sifted through vast data from the Transiting Exoplanet Survey Satellite (TESS), resulting in the identification of approximately 8,000 new eclipsing binary systems.

What are Eclipsing Binaries?

Eclipsing binaries are fascinating stellar systems where two stars orbit one another, with one occasionally blocking the other's light from our vantage point. This phenomenon can be detected by observing dips in brightness of the stars, similar to how exoplanets are discovered. If a star’s light dips significantly, it likely indicates that another star—or a faint companion—is crossing in front.

Harnessing Data Like Never Before

The data from TESS, which monitors a staggering 98% of the night sky, was meticulously pre-processed before being handed over to volunteers. Researchers limited the dataset to stars brighter than magnitude 15 and employed an innovative Python tool known as the ELEANOR pipeline to generate millions of light curves. These curves were standardized to ensure that periodic signals from TESS did not masquerade as eclipses.

AI Joins the Fray!

Once pre-processed, the data was analyzed using a specially trained convolutional neural network. This AI was designed to recognize eclipse shapes rather than just periodic signals, enabling it to efficiently identify eclipsing binaries, successfully locating about 85% of known systems from TESS and roughly 56% from Kepler. Surprisingly, it also discovered around 32% of exoplanet candidates within TESS's dataset!

Crowdsourcing the Search

Even after rigorous AI analysis, the dataset was not ready for public consumption. The research team, alongside a select group of trained volunteers, utilized a platform called Exogram to curate a list of 10,000 candidate targets. These were subsequently made available on Zooniverse, a crowd-based research site, where from September 2024 to March 2025, volunteer citizen scientists completed an astounding 320,000 classifications of eclipsing binaries.

Astounding Results Unveiled!

This monumental effort culminated in the discovery of 10,001 eclipsing binary systems, with 7,936 being completely new to science! The remaining 2,065 systems were already known but received revamped parameters thanks to the insights gathered from TESS. Some of these systems displayed intriguing characteristics, including variable eclipse timings and potential third star companions, waiting for further exploration.

The Future Awaits: Exoplanets on the Horizon

As exciting as these discoveries are, they hint at an even greater quest. TESS was originally designed as an exoplanet hunter, and the collaborative analysis methods used in this study offer a promising pathway to deeper insights into exoplanets. Dr. Kostov expressed his eagerness, stating, "I can’t wait to search them for exoplanets!" With a team in place and data ready for scrutiny, we can anticipate groundbreaking findings on the horizon.