The raw and the cooked (repost)

(this material was originally published September, 2018 at

“Raw data” is both an oxymoron and a bad idea.

– Geoffrey C. Bowker, Memory Practices in the Sciences

“What you will see from these presentations is policing at its rawest,” said LAPD Chief Charlie Beck at a press conference about the agency’s new Critical Incident Video Release practice. In their first such presentation, LAPD Public Information Director Josh Rubenstein describes Critical Incident Video Releases as offering a comprehensive anthology of evidence from investigations in progress:

“Our community briefings will show you physical and digital evidence, including photographs of the scene, pictures of any weapons recovered, audio recordings of 911, radio transmission recordings and video footage from our body-worn and digital in-car video system.”

Critical Incident Video Release NRF031-18 was released to the public via the Los Angeles Police Department YouTube channel on June 20, 2018. It showed footage from a May 6 encounter between a 24-year-old man and LAPD officers in the department’s Newton Division, near Exposition Park. The video release was praised as a step forward for a police department that was quicker to adopt bodycams than it has been to make the resulting video recordings available to the public. But the video was also critiqued for its slick production values, fragmentary depiction of the event, and potential to only show officers in the best possible light.

The NRF031-18 video itself, as well as the new policies that led to its creation and release, illustrates several points raised by the Community and Culture working group at the OTRATT National Forum meeting in 2016. Among them are the difficulty of controlling how people perceive video (both as an objective, neutral record of “what actually happened,” and as a record that can be ambiguous, inconclusive, or easily misread) and the newly urgent need for media literacy on the part of both public information officers and public viewers of police video. NRF031-18 also demonstrates how difficult it will be to make ethical, effective new video evidence policies that support increased transparency and accountability for law enforcement.

First, let’s talk about the content of NRF031-18—which, considered quantitatively and critically, makes a pretty weak case for showing “policing at its rawest.” The 17-minute release contains only about 4.5 minutes of bodycam video and 911 call audio. That’s only 0.15% of the “nearly 50 hours” of video reportedly collected in relation to the incident, (most of which will likely never be released to the public). The minute or so of 911 call and police radio dispatch recordings seem to be substantially complete; at least, no edits are apparent to the ear, and none are referenced onscreen or in the introduction to the audio. This impression of completeness is further reinforced by continuous on-screen transcription of the calls over a visual background of uninterrupted—but almost certainly unrelated—audio waveforms.

The transcript of 911 call audio appears on a background of audio waveforms. The fine blue line (just between the W and h in “What’s”) travels from left to right as the audio plays. Mismatch between the audio playback and the waveforms shown suggests that this image is likely generic visual “wallpaper,” not the waveforms of the actual recording.

The body-worn camera footage, however, is heavily edited, with gaps between segments ranging from a few seconds to several minutes in length. Some of these cuts are indicated by onscreen text; others are described verbally by Alan Hamilton, a Commander in the LAPD’s Force Investigation Division, whose narration and framing of the edited bodycam video segments comprises most of the remaining 12 minutes.

Some edits in the bodycam video footage are signaled by onscreen captions; others are described in the narration.

People are, by nature, credulous about what they do see, and suspicious of what they don’t. Police may redact footage or withhold it from public release for good reasons: to protect subjects, to ensure due process. But withholding and redaction also, inevitably, fuels speculation. An unexplained gap of a few seconds is enough to spark conspiracy theories about “what THEY don’t want you to know.” This tendency to obsess over the unseen is present with all forms of redactable evidence, but it’s particularly potent with video evidence—and we’ve actually known that for decades. In their 1979 book Videotape on Trial: A view from the jury box, Gerard Miller and Norman Fontes discussed results from several studies examining the use of taped testimony and trials and their influence on juries:

“We observed that when jurors knew material was edited, they speculated about its content, an activity that might be even more biasing than knowing what the excerpt contained and being instructed to disregard it.”

Were those missing 30 seconds of bodycam video cut because they were unremarkable—a mere continuation of the exchange as we had already seen it? Or maybe because a local resident, not involved in the incident, happened to walk in and out of the camera’s view, and cutting the segment in which they appeared was the easiest and most effective way of protecting their privacy? Or perhaps because the officer or the man he was addressing used profanity? As it is, we can only speculate; however, the briefest of onscreen notes about what the excerpted material contained could smooth over that gap very effectively, while providing additional transparency about departmental policies on redaction of video prior to public release.

This clip is identified in the upper right corner as being from the passenger officer from the first car to arrive at the scene. Other clips show different ID camera ID numbers, and the same camera ID also appears without the “PASSENGER OFFICER” notation.

Next, let’s look closely at what happens within the video frame. The first segments of bodycam footage shown are verbally and visually identified as being from the passenger officer in the first patrol car to respond to the 911 call. While Hamilton does not ever mention that footage from any other officers’ body-worn cameras is being shown, the ID numbers visible along with the embedded timecode in several other video segments in the edited package suggest that viewers are seeing footage from at least three different bodycams.

This clip does not include the timecode area or camera ID in the upper right corner, and may have been cropped from the full-frame video image (see below).

Several other segments of video have the timecode area cropped out entirely. This may have been done to focus more closely on areas of interest within the frame, but the cropping itself is never acknowledged in narration or with onscreen text, and the visual quality of the cropped video is consistent enough with the uncropped images that a casual viewer easily may overlook the absence of the embedded timecode and ID numbers.

Above: Uncropped, full-frame video from a second officer’s bodycam. Below: Comparing the corresponding area of the full-frame image (left) with the cropped image (right), and the relative positions of features like the officer’s gun butt and the palm tree trunk in the left foreground, suggests that they’re from the same camera source.

And speaking of timecode: None of the camera clocks are set to the local time (PDT). While the times shown on different camera views are fairly close to one another, the individual cameras’ timestamps do not appear to be synchronized. Events described as occurring earlier in the chronology of the event are marked by at least one camera with a later time than events that occur later in the narrative and are shown from a different camera view. Although most of the narration describes the events in sequence and with fairly precise timing (“six minutes later,” “after approximately 32 minutes of trying”), Hamilton also says that “at some point during the interaction” Chavez picks up a flower and holds it out toward the officers. Inconsistency among timecodes makes it much harder than it should be to confirm that the released video clips fits a clear event timeline; by the same token, the specificity possible when working with timecoded digital video makes any vagueness or ambiguity in the narrative about what happened when far more noticeable.

Non-incident footage that appears in the video includes officers putting on their chest-mounted bodycams and images of the less-lethal force weapons used by LAPD.

Turning to the narration used throughout much of the piece, it becomes clearer how most of NRF031-18’s impact as a tool for transparency comes not from the bodycam footage itself, but from framing and interweaving that footage with specific details about the technology and procedures being used by police in this incident. Hamilton notes that the two officers who responded to the 911 call were both equipped with chest-mounted bodycams. Later, he describes the “low ready position” the passenger officer uses when he draws his gun, and reiterates that the chest-mounted camera perspective might create a misleading impression of where the gun is pointed. Hamilton describes, and the video shows, the Remington 870 shotgun and beanbag rounds used in the department’s “less lethal” force protocols. As the officers close in, Hamilton describes the “arrest team” and enumerates its members’ designated roles (from controlling arms and legs to using lethal force, if necessary), and explains why the officers use two sets of handcuffs to restrain the man after they Taser him and bring him to the ground.

That narrative is not strictly informative, though. It’s also descriptive of the footage to come: “You are going to see him try to communicate with [the suspect],” “you may not be able to see the actual angle of the officer’s weapon is not directed at the suspect’s body at this point,” “here you’ll see him spit in the direction of nearby officers.” It’s Presentation 101—tell them what you’re going to show them, show them, then tell them what you showed them—but it’s also Persuasion 101—tell them what they’re going to see, and you might convince them to see it. This approach can create the impression that the footage can’t simply speak for itself. (The tradeoffs inherent in presenting ambiguous or hard-to-decipher footage, with or without comment, are something to address in a later post.)

Visitors to online news outlets that covered the Critical Incident Video Release are directed to the LAPD YouTube channel to watch the video. It can no longer be viewed as an embedded feature.

Finally, a few thoughts on the mechanism of the video’s release, and the ethical dimensions of that decision. NRF031-18 didn’t drop as hard as the surprise release of THE CARTERS’ Everything Is Love the previous week, but local and national press outlets still covered it and featured the LAPD video in online stories. Those links no longer work as embedded features of the news stories, though; instead, potential viewers are redirected to the video on LAPD’s YouTube channel, where comments on this video have been disabled. Steering viewer traffic toward a single, verifiable source (rather than allowing clips to auto-play directly from embedded links) and disabling comments at that source helps contain the spread of violent and disturbing imagery. It also greatly minimizes the amount of politically-charged rhetoric generated by the video and about its subjects, and permanently associated with it online.

An LA Times story about NRF031-18’s release, on the other hand, did have comments enabled. The three comments that were online at the time of this writing included critiques of the edited nature of the footage shown; one, in particular, echoes OTRATT contributor Mary Fan’s proposal of a “bounded access model” for proprietary knowledge that balances private-sector interests with public safety:

“Maybe it would be better if a neutral third party did the editing so that LAPD couldn’t be accused of bias in deciding what to keep and what to cut.”

A truly neutral third party would likely be difficult to find in any city — let alone one with such long-standing tensions between police and the public, and with so many experienced film editors looking for work. However, the OTRATT Culture & Community group did also question whether the public might see a meaningful difference between civilian and sworn personnel’s stewardship of video evidence—and whether other city agencies with a higher level of public trust, such as municipal archives or public libraries, might effectively collaborate with police departments to provide reliable preservation and access to evidentiary media.

The group also discussed how core competencies of the public information officer in the bodycam era now include critical media literacy—that is, the ability to understand the complex nature and many limitations of video evidence, and video as a format. Public information officers must pay close attention to what material they’re putting out, and how; they must be attuned to how standard operating procedures and accepted practices come across on camera; and they must be able to anticipate how editing or redacting footage (regardless of its source) will impact public perceptions of trustworthiness and authenticity. NRF031-18 was heralded as a major, if early, step in the LAPD’s effort to increase transparency. This is fair enough; a major city police department releasing any case-related video within just a few weeks of a use of force incident that ended in a man’s death really should contribute something to that aim.

But we must also read this carefully crafted package of clips and context as an attempt to shape truth perceptions, as well as a necessary move to better inform the public about specific aspects of law enforcement practice and procedure. If LAPD (or other departments who take this approach over releasing unedited or minimally redacted footage) are not sufficiently transparent about their joint aims of making incident records public while teaching viewers “more about the department and policing as a whole,” they will undermine both efforts significantly. Like so much other data, the NRF031-18 video is raw only in the sense that a sashimi platter is raw: it presents for consumption a lot of little pieces that were selected, cut up, and artfully arranged by chefs who are doing their best to avoid any heat. At some point, if you do enough slicing, you really have to call it cooking.