The independent news organization of Duke University

You asked about the Duke data set study. Here's what we know so far

Where cameras were located in March 2014 for the DukeMTMC database study. Open Data Commons Attribution License. Courtesy of Megapixels. Source: Ergys Ristani, Francesco Solera, Roger Zou, Rita Cucchiara and Carlo Tomasi.
Where cameras were located in March 2014 for the DukeMTMC database study. Open Data Commons Attribution License. Courtesy of Megapixels. Source: Ergys Ristani, Francesco Solera, Roger Zou, Rita Cucchiara and Carlo Tomasi.

Last week, The Chronicle reported that thousands of pedestrians on Duke's campus were recorded as part of a study by Duke computer science professor Carlo Tomasi. The researchers then used the movements and likenesses of people captured on camera to create a data set, which was publicly disseminated and has been used by professors and military researchers around the world.

Tomasi has since written a letter to the editor clarifying his lab's research objectives and apologizing to those who were recorded.

The Chronicle asked our readers what they would like to know about the story, and our reporters are currently searching for answers. Feel free to submit your own questions at the bottom of this page. 

Here's a sampling of the questions you wanted answered and what we've been able to find out so far.

Q: "Are there any pictures of the alleged posters that were set up to notify pedestrians about the filming? Specifically where were they posted in relation to the filming locations?"

A: The posters were placed "at all entrance points to the recording area," according to Tomasi. Recording areas included the open space immediately in front of the Chapel, the areas at the entrance of Page Auditorium and the Divinity School, the walkways on the sides of the Chapel and the bike rack and sidewalk on the right side leading up to the Chapel steps. 

Imprinted on the posters was a description of the research and the researchers' contact information in case any passersby wanted the data to be erased. According to Tomasi, no one contacted the researchers asking to be exempted.

The Chronicle has not been able to access an image of the posters. 

Q: "How can current and former students, faculty and staff determine whether their face was included in the data set?"

Q: "I work in the library at Duke. How can I determine whether my image was captured and, if it was, whether has been used?"

Q: "I was in my final semester at Duke when this info was collected. Is there any way to find out if my likeness was captured?"

A: The Chronicle has not been able to obtain a list of the people whose images were captured and used in the data set, nor do we know if such a list even exists. The list may not exist because the study was intended for multi-camera tracking and didn't require anyone's names. Nonetheless, we will report on this issue as we learn new information. 

Q:  "I was an undergraduate student during March 2014, and I'm very curious about why this was done. Is there any way to check if we were photographed? Why did they do this at Duke? What are they doing with this information in all of these papers? The IRB was for an indoor space... how was this legal?"

A: The authors of the study have not explicitly stated why the research took place at Duke, but the paper where they present their dataset offers some clues. The recording was scheduled to take place at class change times "when pedestrian traffic is heavy," which provided more subjects to analyze.

The paper also describes the recording as taking place in a natural environment using eight wide-angle cameras, with the main challenge being "persistent tracking under occlusions and blind spots." The researchers wanted to simulate the real world to challenge computers—passersby are carrying objects such as backpacks or umbrellas, some are riding bicycles, and the paper notes that people often loiter in blind spots.

It seems that the quad in front of the Chapel offered an area where people were freely and naturally moving and thus a wide-open space easily viewable by cameras. Three of the five co-authors of the paper were affiliated with Duke at the time, as well.

Recording outdoors went against the guidelines set forth by the Institutional Review Board, according to Michael Schoenfeld, Duke's vice president for public affairs and government relations. 

Regarding the question of legality, Schoenfeld wrote in an email to The Chronicle that outdoor campus areas and open events such as games or concerts are public spaces. This means that anyone in these areas can have their photo captured, whether it be for an Instagram post or by a security camera.

We are currently looking into the disconnect between the IRB and the researchers, how this information was used in subsequent papers and whether there is any way for individuals to find out if they were recorded.

Q: "Will the Duke investigators face sanctions or otherwise be held accountable to what appears to be a gross violation of an IRB-approved research protocol? Any comment from IRB officials?" 

A: The Chronicle is actively pursuing this line of investigation and will publish a story as soon as we obtain more information. We have yet to receive any comment from the IRB.

Q: "Were participants in the research offered a chance to sign consent forms? Were Durham residents, and others visiting the campus, included in the research?"

A: Posters were displayed at entrance areas to the recorded areas near the Chapel with the researchers' contact information if individuals wanted to have their data deleted, according to Tomasi. He made no mention of consent forms.

The paper describes the capturing of more than 2,000 "different identities," or "distinct persons," on film and does not further characterize who was recorded or if the researchers took note of who was in the videos.

Do you have a follow-up question or any information that might answer one of the questions listed here? Let us know using the form below:


Share and discuss “You asked about the Duke data set study. Here's what we know so far” on social media.