Selling School Safety
- greta914
- Sep 3
- 22 min read
Updated: 10 hours ago
Corporate Marketing of AI-Powered Vape Detectors as Educational Technology
Authors
Todd Whitney, Independent Researcher-Investigator/ Affiliate, Present Moment Enterprises*
Yung-Hsuan Wu, Independent AI Auditing Researcher; Affiliate, Present Moment Enterprises*
Greta Byrum, Principal Owner, Present Moment Enterprises*
Eloise Gabadou, Independent AI Governance Researcher
Clarence Okoh, No Tech Criminalization in Education (NOTICE) Coalition
Marika Pfefferkorn, Twin Cities Innovation Alliance (TCIA) & NOTICE Coalition
*Main contact authors
Introduction: Selling School Safety via Surveillance Tech
As the US educational system embraces advanced tech solutions to manage safety, schools are installing growing volumes of sensors, cameras, and other data collection and processing systems. These systems offer a vision of technology-enabled efficiency to solve the many complex issues faced by the education system in recent years -- school shootings, disease transmission, widening income and resource gaps, and more.
Meanwhile, backed by venture capital, tech companies are expanding their reach into new markets, including the educational technology (“edtech”) market, adding data extraction and processing layers to existing technologies to capitalize on the popularity of artificial intelligence (AI).
HALO, a “smart sensing” company under the Motorola Solutions umbrella, is a leading vendor of vape detection technologies (VDTs). Its devices and systems are marketed as answers to a range of issues in schools, including air quality concerns, airborne disease transmission, and student use of tobacco and marijuana e-cigarettes.
Vape Detection Technologies (VDTs) are AI-powered systems increasingly used in US schools to monitor bathroom environments for e-cigarette vapors, prohibited substances, and behaviors that may indicate vaping. VDT systems comprise arrays of sensing devices installed in school bathrooms, locker rooms, and other “private” spaces in schools, programmed to collect and feed data into an AI-driven backend platform for analysis. While VDT systems do not collect visual data, manufacturers claim the devices have the ability to monitor for aggression and bullying, in addition to air quality and disease transmission in schools using built-in sensors. The systems also allow school officials and law enforcement to cross-reference device alerts against data from hallway CCTV, electronic hall passes (E-Hall Pass), and other data to circumstantially identify students allegedly involved in vaping incidents.
VDTs are sold as all-in-one solutions, including not only hardware devices including sensors (called “snitch pucks” in WIRED’s recent reporting on VDTs), but also a software-as-a-service (SAAS) cloud-hosted monitoring platform that digests data and spits out recommendations for student disciplinary action. With these purported solutions, Halo and other VDT providers invite schools to outsource and streamline disciplinary systems via proprietary infrastructure, ingesting student data into its cloud servers to generate trend analyses based on continuous monitoring of private spaces.
And yet, under examination, it is not evident that VDT systems are effective or successful at achieving all or even any one of the many promises made by the marketing literature. In the following pages, we analyze a series of HALO’s marketing claims based on data and subject matter expertise.
A note on HALO: The commercial vape detection market consists of a range of private companies and third-party intermediaries, each claiming to possess slightly different technical capabilities. While our analysis has broader implications for similarly situated VDT systems, this report specifically examines the claims offered by HALO (formerly IPVideo Corporation), a subsidiary of Motorola Solutions. We focus the current analysis on HALO for three primary reasons:
HALO is one of the largest providers of VDTs to schools in the country; the company reports that it does business with over 1500 US school districts. Given HALO’s market share, this report’s findings hold implications beyond Minnesota.
This report places a specific focus on the rise of VDTs in public schools in the state of Minnesota as an extension of the NOTICE Coalition’s Automating the School Surveillance report. HALO is the most widely adopted VDT brand in Minnesota schools.
Our research team obtained public records from the Minneapolis Public School District regarding the District’s 2024-2025 academic year VDT pilot program, which utilized HALO technologies. This resource represents the first publicly available dataset on the actual performance of a commercial VDT system in a public school setting.
HALO is in the news: A WIRED article from August 8, 2025, found that “any hacker on the same network could have hijacked a Halo 3C to turn it into a real-time audio eavesdropping bug, disabled its detection capabilities, created fake alerts for vaping or gunshots, or even played whatever sound or audio they chose out of the device’s speaker.”
Taken together, these factors led us to narrow our current focus on the impact of HALO’s products on public schools in Minnesota. Our findings hold implications beyond this specific context, suggesting the need for future research into additional VDT systems.
Examination of VDT Marketing Claims
Claim 1: VDTs can accurately detect student vaping in school settings
“The device has been extremely accurate and has helped school leadership deal with the vaping issue effectively by identifying when and where our students have been vaping. In addition, it has given our school the ability to differentiate from typical vaping to those that have THC in it.” (HALO marketing material 2, testimonial)
Analysis
Review of HALO's materials yields scores of enthusiastic quotations about its effectiveness, but when we looked at data from MPS, we found evidence that the advertised abilities did not live up to the marketing hype.
The alerts are overwhelming and misleading. VDTs produce a high volume of notifications, but this flood of alerts may reflect overly reactive sensing levels or misuse rather than actual vaping incidents. False positives may also occur as a result of humidity changes in the environment, or unrelated gases produced by cleaning agents and aerosolized beauty products. As a result, these systems could create an inflated estimate of vape use by students, as well as a false sense of certainty for administrators, while offering little actual insight.
Data from Minneapolis Public Schools’ HALO VDT systems show that the number of alerts generated by sensors placed in four schools varied between 1000 to 7000 alerts each month – a fluctuating rate of incidents that appears unlikely to align with the reality of student vaping behavior. In the most extreme case, we found that Camden High received more than 600 alerts in a single day — which translates to more than one alert per minute during school hours demanding school administrators' attention. (Fig. 1).
Too Many Alerts. We received and analyzed nearly 7 months’ worth of vape detector alert data from four Minneapolis public schools (Anderson Middle, Anwatin Middle, Camden High, and Roosevelt High). (Fig. 1: Camden High - Monthly Alerts and Average Daily Alerts, September 3, 2024-April 15, 2025.) |
Claim 2: VDTs enable accurate environmental health monitoring in schools.
“Measuring AQI is determined by the EPA. It is used to evaluate the quality of the environment regarding the chemicals in the air. An AQI reading can provide us with a number that represents how polluted the air might be. The HALO Smart Sensor provides an AQI measurement that is a rolling average of the quality of the air you are breathing over the course of a few hours.” (HALO marketing material 3)
Analysis
HALO markets its devices as air quality monitors, but misrepresents the basic scientific and regulatory standards expected of environmental health tools.
In the United States, there are no legally enforceable indoor air quality (IAQ) standards. Organizations like OSHA offer guidelines for indoor work environments, and the EPA publishes outdoor air quality indices, but there is no federal benchmark for evaluating complex indoor environments like schools. HALO’s marketing exploits this regulatory gap by presenting its devices’ outputs as conforming to regulatory health safeguards, without the burdens of actual compliance or independent validation based on acknowledged standards.
The data HALO provides is not based on EPA standards.While HALO devices report measurements on pollutants like NO₂, VOCs, or Particulate Matter, it does not use the EPA’s Air Quality Index (AQI) algorithms to determine thresholds. Instead, HALO leverages the resemblance of EPA regulations while using simplified standards to give the air quality a “score” on its proprietary index. Further, systems administrators have the ability to set custom thresholds for a variety of air quality standards, raising questions about HALO’s AQI Index as a tool that reflects scientifically sound air quality risks.
These false claims undermine both the credibility of the device as a public health tool and HALO’s claims regarding accuracy.
Good air quality is an important foundation for any learning environment. Studies prove that degraded air quality in schools leads to decreased student performance, higher absenteeism, and difficulty concentrating. Schools looking to improve indoor air quality deserve transparent, validated, and standardized approaches to measuring the performance of products in the indoor environment.
Devices that direct a school’s HVAC system responses should ensure optimal sensor accuracy, placement, and on-going calibration to guarantee that any inferences drawn are meaningful and productive. Further, these devices should provide school leaders with actionable, understandable data that can guide ongoing ventilation adjustments to address air quality issues as they change over time. HALO projects these capabilities in its products, but does not provide information or data on the effectiveness of its devices in creating better air quality environments within schools.
Environmental health monitoring is not just a bonus feature.
Calling a multi-sensor surveillance device an air quality monitor is misleading if it doesn’t utilize scientifically grounded standards or provide actionable data. Worse, it may give schools a false sense of environmental safety while distracting from more effective interventions like CO₂-based ventilation strategies, HEPA filtration systems, and the EPA’s Indoor Air Quality Tools for Schools framework. Instead, HALO’s IAQ metrics are often bundled with other features (vape detection, aggression monitoring, etc.), diluting its legitimacy as a health-focused technology.
Claim 3: VDTs can detect and limit disease transmission
“The HALO Smart Sensor Health Index provides a real-time indication of the potential risk for the spread of airborne infectious disease in a building.” (HALO marketing material 4)
Analysis
The HALO Health Index does not measure, track, or limit the transmission of infectious diseases.
Despite marketing language that suggests HALO devices can help mitigate disease spread in schools, none of the sensors included in HALO’s system are capable of detecting pathogens or transmission events in real-time. Instead, HALO’s Health Index measures Carbon Dioxide (CO2), Particulate Matter (1 μm, 2.5 μm, 10 μm), Relative Humidity (RH), Total Volatile Organic Compounds (VOC), and Nitrogen Dioxide (NO2). While these measurements may be indirectly relevant for inferring potential disease transmission and provide a rough series of recommended actions, HALO’s Health Index cannot offer insights on the presence of bioaerosols in any direct or medically meaningful way.
HALO’s Health Index is a marketing construct, not a medically validated tool.
HALO frames its Health Index as a real-time indication of disease transmission risk, but the metrics it uses only focus on a limited set of pollutants and environmental variables such as volatile organic compounds, carbon dioxide, temperature, and relative humidity. These variables may provide useful background information about indoor comfort and ventilation, but they do not directly measure pathogens or infection risk.
Scientific research shows that temperature and humidity can influence how long viruses remain viable in the air, and that elevated CO₂ may signal poor ventilation. However, none of these factors alone, or in combination, offer a reliable or standardized way to quantify infectious disease spread.
Further, evaluating the risk of airborne disease transmission must account for the built-environment, occupancy, and other variables when validating the efficacy of interventions and recommendations provided by the HALO Health Index. A standalone Health Index score representing airborne risks may actually downplay the complexities and context-dependent nature of airborne infectious disease transmission. This makes HALO’s claims that its “Smart Sensor Health Index” is a tool that “provides a real-time indication of the potential risk for the spread of airborne infectious disease”, speculation.
By presenting its proprietary Health Index as a health protection tool, HALO risks misleading schools into a false sense of safety. Reducing airborne transmission requires validated strategies such as air filtration, ventilation, calibrated monitors, and development of a building readiness plan for potential transmission events. Schools deserve tools that support transparent, evidence-based decisions, not indices and risk scores that conflate comfort metrics with potential medical outcomes.
Claiming health protection without medical validation may mislead school districts and other consumers.
HALO has actively promoted this health framing in pursuit of public funding, including by helping schools apply for school safety grants and other funding opportunities targeted at health and air quality under the guise of infection control. But improving ventilation and monitoring pathogen risk requires specialized instruments and informed administrators, not generic environmental sensors bundled with behavior surveillance features.
Claim 4: VDTs can detect aggressive behaviors
“HALO uses machine learning to recognize abnormal noise patterns. It learns normal sound levels and alerts when noise exceeds set thresholds. Advanced analytics enable accurate aggression detection.” (HALO marketing material 5)
“HALO detects distress keywords and aggression in real-time, providing comprehensive security with a single device. It covers up to 1,960 square feet, ensuring broad protection in any space.” (HALO marketing material 6)
“HALO alerts school staff to aggression and bullying through spoken keyword and sound level recognition. HALO stands out among other vape detectors for schools by offering an enhanced layer of safety.” (HALO marketing material 7)
Analysis
The use of machine learning algorithms to detect aggression, conflicts, and bullying is speculative. Behaviors cannot be interpreted simply through acoustic analysis of sound levels and noise patterns.
Leveraging machine learning to automatically detect verbal aggression in human voices has been a long-standing research problem in affective computing and surveillance technologies. Nevertheless, the scientific foundation that HALO and other microphone-equipped vape detector providers gesture toward in their marketing is partial and inconclusive.
Audio cues are not reliable predictors of aggression.
Verbal aggression detection assumes that machines can identify distinctive vocal patterns that indicate aggressive behavior. Academic studies of “affective science” often trace back to Swiss psychologist Klaus R. Scherer’s work in 1986 and 1991 to show that emotions correlate with vocal features like pitch, loudness, and tempo. While scientists and developers speculate that it is possible to extract these features from a stream of audio, the same vocal patterns can come from multiple emotions. Fear, anger, and joy all produce higher-pitched, louder, and faster speech.
AI developers who claim that their systems identify vocal features linked to fear and anger as proxies for aggression often omit that extreme joy produces similar vocal patterns. They also do not distinguish between ‘hot anger’ and ‘cold anger’, which leads to different vocal performances. This flaw was demonstrated when ProPublica journalists tested an aggression detection algorithm developed for a similar purpose by another company Sound Intelligence: students excitedly cheering for pizza triggered the aggression alarm.
The interpretation of aggression is highly contextual and requires more nuance.
HALO claims to detect and prevent conflicts in multiple settings, such as in corporate offices, schools, or hospitals. However, verbal aggression and aggressive behaviors manifest differently in complex social environments: chanting and shouting in a gymnasium may be friendly and non-aggressive, while the same behaviors in a quiet classroom may indicate tension and aggression.
Given that the interpretation of aggression depends on the context, it is not reasonable to claim that detectors can pick up aggressive signals in all contexts. In the same ProPublica case, journalists found that Sound Intelligence’s system, which was originally tested with audio data collected in the ‘pub district’ of the Dutch town Groningen, performed poorly with audio clips collected from US school kids. Researchers further performed reverse-engineering and found the system was biased toward associating aggression with rougher vocal tones and strains while ignoring high-pitched shrieking.
HALO’s limited keyword spotting capabilities do not help contextualize aggression incidents.
HALO markets its devices as capable of detecting distress keywords to alert staff about aggressive incidents; however, HALO provides no evidence about how well its device works in the context of a school. The detection and recognition of spoken words in a noisy environment is both technically difficult and computationally expensive (i.e., requiring a lot of processing power). Commercial products such as Apple’s Siri that leverage similar keyword spotting techniques often require learning from a distinct user’s voice by asking the user to record themselves repeating a few phrases. HALO devices do not have the capability to learn from tens of hundreds of students and school staff that pass by its sensors each day.
Furthermore, the computational requirements for recognizing all existing words in human speech, different accents, and different languages largely exceed the microprocessing unit that can be installed on the size of the HALO device. In reality, HALO devices could only detect one out of the pre-configured keywords on this list provided in their Administrator Manual:
Help Emergency
Help Nine One One
Help One One One
Help Triple Nine
Help Triple Zero
HALO Privacy
HALO Clear
HALO Check In
socorro auxilio [Portuguese]
aidez moi au secours [French]
aiuto soccorso [Italian]
ajuda emergência [Spanish]
This limited list cannot fully capture the potential range natural speech of students or staff members in distress or when facing an aggressor.
HALO devices can neither detect nor prevent bullying.
Based on these shortcomings in current systems, aggression detection research has converged on the ideal of developing multi-modal mechanisms that consider audio, visual, and contextual cues along with keyword spotting to retrieve semantic information (what is said), potentially combined with camera footage and real-time speech recognition to identify aggressive behavior. Even if such a speculative system were able to interpret context and produce accurate analyses – which is far from demonstrated – HALO devices alone have only a small fraction of these capabilities.
HALO devices do not have the capabilities to capture the context in which ‘aggression’ and ‘bullying’ may arise; in fact, their false alarms based on out-of-context audio cues may only lead to the policing of student behavior, inviting school administrators and security officers look at students through the lens of ‘aggression’. This creates tension and undermines the trust between school teachers and students, which further decreases students’ sense of security under AI-powered surveillance.
Claim 5: VDTs can accurately detect and locate gunshots
“HALO gunshot detection devices use advanced algorithms to analyze sound waves, accurately distinguishing shots from other loud noises by assessing specific acoustic signatures and patterns.” (HALO marketing material 8)
“While many other devices require multiple sensors to achieve full coverage, a single HALO Smart Sensor can cover an entire room up to 1,960 sqft.” (HALO marketing material 9)
Analysis
HALO’s gunshot detection system is unlikely to be precise and cannot pinpoint location.
Acoustic gunshot detection systems, particularly for outdoor police surveillance, are a major focus in surveillance research. SoundThinking’s tool (formerly ShotSpotter), a system widely adopted by US police departments, remains a contentious case study: Brooklyn Defender Services discovered that 99% of ShotSpotter’s gunshot alerts did not lead to the discovery of guns or to people involved in gun violence in New York City. An 8-year study further revealed that ShotSpotter failed to reduce firearm homicides and arrests. Finally, the tool was shown to be significantly less efficient than citizen-initiated calls for services in helping the police identify criminal activities.
While HALO’s Smart Sensor is designed for indoor usage, its gunshot detection capability shares similar issues that outdoor detection systems face, including locating gunshots and differentiating shots from loud sounds like glass shattering, cars backfiring, or doors slamming. Furthermore, indoor detection faces challenges germane to the system’s setup environments.
HALO gunshot detection may not reliably perform in school environments.
HALO claims to be able to identify the unique audio cues that come with a shot being fired; nevertheless, HALO does not report the accuracy and precision of its system anywhere that we could find. In fact, there are currently no standardized tests to certify gunshot detection systems in their operational environment. Most studies are based on either computer simulations or lab-environment testing, and many are based on recorded audio clips of gunshots and noises, which do not capture the real-world ambient environment of a bustling school. Others proposed using multi-modal verification mechanisms, such as infrared or visual object recognition, to enhance detection performance; these are not capabilities that HALO devices possess.
HALO’s claims to detect gunshot locations are misleading.
For a gunshot detection system to precisely locate where a gun is fired, triangulation via multiple sensors placed in different locations is required. HALO’s claim that it can cover a large space with discrete devices in bathrooms is misleading, given that schools are made up of multiple enclosed spaces with furniture obstructions, walls, glass, and doors instead of one empty hall. The presence of sound-deadening fabrics in walls and doors may affect the clarity and range of detection; sensors may pick up on shots fired in another room other than the one it is installed; room reverberation may create overlapping sounds, which require a special array of microphones to mitigate. HALO’s claim underestimates these environmental complexities and oversells the devices’ capabilities.
Claim 6: VDTs make students feel safer and protect privacy
“Our decision to select HALO was based on the versatility of the sensor and the positive reviews from other schools. We made the right decision deploying HALO within our school district as it’s protecting our school privacy areas and also providing security while protecting individual privacy.” (HALO marketing material 10, testimonial)
“Students have told me they finally feel comfortable using the school bathroom again. That’s really our heart and goal in the issue, is to prevent vaping, not to try and catch more kids making poor choices.” (HALO marketing material 11, testimonial)
Analysis
Evidence suggests that HALO VDTs undermine students’ sense of safety and privacy.
Students have been taking to social media platforms like TikTok and Instagram to share their thoughts about HALO devices being inaccurate and easily triggered. Some examples include: “if your perfume is too strong it can set it off,” “my school can’t do this as axe [body spray] also sets them off and the teachers vape in the bathroom,” or “if like a bunch of ppl walk under it it’ll say ‘particular levels exceeded.’”
(Screen capture from social media*)
*HALO devices were mentioned specifically either in the picture/video or in comments related to quotations reviewed for this report.
Other students mention HALO devices not working: “bruh I think they were fake or something, they never went off 😭” “Literally I go to hs and the one we have don’t work at all.”
Students report perceived overreaction to alerts: “Same bro my principal and vice principal both came into the bathroom and said they got alerted from the smartwatch that the smoke detector went off,” “I noticed it start lighting up so I left but when I was walking back to class I saw security going towards the bathroom 😭.” This could not only reinforce the perception that the school is overly securitized (“They search everyone in the morning 😭”) but also result in an inefficient use of resources.
Placing surveillance-capable devices in restrooms undermines student privacy and raises ethical concerns. HALO markets its devices as “privacy-safe” because they don’t record audio or video, but the reality is more complicated. These devices contain sensors that collect high-resolution/high-timeframe environmental data like changes in chemical composition, sound thresholds, body heat, and motion, along with audio data. These metrics are used to identify and infer information about individuals in sensitive spaces. When combined with access to other surveillance technologies like CCTV, administrators can make assumptions about what is happening without consent from monitored students.
VDTs transform safety tools into surveillance systems.The features that make VDTs “smart” (their backend dashboards, live alerts, and detailed logs, etc.) create the potential for misuse, misunderstandings, and overreach. Anyone with administrative access can observe, analyze, and act on these sensor-triggered events, often without clear oversight or accountability. In practice, this can mean real-time digital monitoring of students in places meant for privacy.
Restrooms are not neutral spaces.Bathrooms are often the only places students have to decompress, take a moment of silence, or get away from overwhelming classroom environments. When students know they are being monitored, they may avoid using the restroom altogether, raising concerns around mental health, medical needs, and basic dignity.
Do VDTs reduce student vaping?
Overall, we found no publicly available evidence of a direct link between the use of HALO’s VDT systems and consistent, ongoing reductions in the rate of student vaping. HALO provides testimonials from its customers, but no replicable studies or data showing change over time in places where its systems are in use.
Without independent research-based evidence, the many questions about validity and accuracy of HALO’s devices, systems, and capacities raise serious concerns about the use of these technologies, especially without regulation or safeguards.
Moreover, nothing in HALO’s literature or marketing claims addresses the need for supportive public health approaches to vaping cessation, nor the underlying factors that may bring students to use e-cigarettes. If schools take the path of procuring “streamlined” solutions that center disciplinary action as the primary method of addressing vaping behaviors, while also outsourcing disciplinary decisions to third-party technology providers, underlying issues will not be addressed.
Critically, VDTs do not address why students vape. Young people cite reasons such as anxiety, depression, boredom, family influence, or simply trying to cope with a difficult world. “Because of how f**ed up our world is,”* said one student; others noted “to feel less anxious,” or “because their parents or older siblings do it.”
Effective responses focus on root causes, not just surveillance. VDTs may give administrators a sense of action, but students overwhelmingly report that they “literally stop nobody.” Building a safer school environment means engaging with students' real needs and not just policing their behavior.
What’s Next?
“Vaping detectors are integral in giving students their bathrooms back. These devices impede students from vaping in school restrooms without getting caught. Therefore, instilling a sense of safety and privacy for students when they are using bathrooms at school.” (HALO marketing material 1)
Earlier this year, the NOTICE Coalition and Twin Cities Innovation Alliance (TCIA) engaged in community consultation with key stakeholders on the impact of VDTs in public schools in Minnesota. Early feedback from those conversations indicates that educators experience increasing pressure to report on student activity, and are held responsible for generating observational information about students to explain the data generated by VDTs. This pressure creates a cycle in which the very adults responsible for social/emotional learning are also asked to report on students in ways that may cause young people to be caught up in a web of outsourced disciplinary action, breaking trust and limiting the ability of responsible adults to exercise independent judgement about student behaviors.
In fact, VDTs appear to reduce students’ sense of privacy and safety, creating an environment of surveillance and suspicion and eroding trust between students and teachers. In previous research and interviews students describe the systems as “reactive, punitive, and ineffective.” Rather than stopping vaping altogether, VDTs tend to shift the behavior outdoors or to other less-monitored areas. As one student commented in a 2025 study from Preventative Medicine Reports, “Then I’ll do it in the classroom.” Even in restrooms, many students adapt by "zeroing" their hits – i.e. holding in vapor so long that nothing is visible when students exhale into toilets, sinks, or shirts. “That’s why I zero it,” one user explained. VDTs may give administrators a sense of action, but students overwhelmingly report that they “literally stop nobody.”
Altogether, systems like HALO’s VDTs depend on the ingestion of data from different sources with varying levels of reliability and sophistication, then digested through opaque machine-driven calculations which make the results appear rational, streamlined, and neat. Without clear external testing and evidence of accuracy and reliability, there is no way to tell whether these systems – which carry an outsized influence on the lives of students – are hallucinating by generating false positives, misidentifying students, or attributing characteristics or behaviors that are not grounded in empirical reality.
VDT marketing materials appear to exploit public misunderstandings of AI capabilities, playing up the mystique that the algorithm is a neutral, effective, sophisticated means to increase school safety. Because the United States lacks comprehensive data privacy legislation, AI technologies like VDTs remain largely unregulated under federal, state, or local law. In particular, there are virtually no independent tests of accuracy or consumer protection standards.
Takeaways
Transparency and governance policies are essential, but missing. The proliferation of high-tech solutions marketed to school systems is outpacing administrators’ and districts’ capacity to establish reasonable and clear policies to ensure that new technology is used properly and safely. These devices generate large volumes of sensitive data – and yet many schools do not publish – and may not have – policies and plans for managing and protecting it.
Schools and other institutions need support to effectively procure, apply, and evaluate the performance of technologies procured for public use, and to create accountable tech policies and guidelines.
Public health should not be a marketing gimmick.
Framing VDT surveillance hardware as a disease control or environmental health tool exploits community concerns about health and well-being without delivering medically useful results. Schools and students deserve better than misleading scientific metrics repackaged as health protections, especially when basic monitoring systems and ventilation strategies are still lacking in many schools, despite widespread concerns raised during the COVID-19 pandemic.
The same applies to exploitation of community concerns regarding school shootings Based on our current analysis, VDTs do not provide meaningful services related to student safety from gun incidents.
Marketing VDTs as solutions to public health problems is dangerous and counterproductive.
VDT systems may place student data at risk.
Information is scarce regarding who has access to VDT-collected data, how it is stored, and how it is used across school systems and law enforcement partners, as well as HALO or other VDT providers themselves and their third-party cloud, hardware, and data partners. Even beyond questions about how schools may use or manage student data, we do not know how or to what extent student data is anonymized in predictive models, which are built from the ingestion of data across its customer base. In the absence of regulation, publicly available policies, and community as well as parental consent, these devices risk becoming part of a broader trend toward ambient surveillance in schools, normalizing the idea that students must be surveilled to be safe.
Research, testing, and benchmarking are needed for VDTs and other educational technologies.
Tech tools procured with public dollars and used to discipline students should be rigorously tested for purpose. In this report, we evaluate a set of claims made by HALO’s marketing materials for which the company provides no independent evidence. As a next step, these claims, and those of other VDT providers, should be empirically tested in both laboratory settings and in situ in school environments.
Without sufficient testing, schools are allowing students to be used as technology test subjects – with neither the ability to verify that funds dedicated to this purpose are being spent responsibly, nor that students are benefiting in any way.
To truly address vaping, provide cessation support.
Currently, the most effective and empirically proven method to stop student vaping is to provide counseling and cessation programs. A holistic approach to address the root cause is needed to understand the costs, consequences, and trade-offs of VDT systems against those of counseling and other traditional cessation methods. Without a scientific approach like this, unproven claims from technology providers like HALO could divert much-needed resources from proven programs to science-fiction solutionism.
Resources
VDT capabilities and Indices: HALO DEVICE FUNCTIONS
Minnesota VDT dataset (github): https://github.com/yunghsuanwu/vape-detector-minneapolis
About this Report
This report is intended to build awareness of the use of VDTs; promote critical examination of manufacturers’ claims of effectiveness; and to propose a set of queries and parameters for VDT hardware and systems testing. Overall, we propose to document and examine the impact of the experimental use of novel data- and AI-driven technologies in complex social settings with vulnerable populations.
Authors & Contributors
The NOTICE Coalition and the Twin Cities Innovation Alliance (TCIA) work in partnership to understand and address the impact of increased surveillance technology use in US schools. In 2024, NOTICE and TCIA published Automating School Surveillance, a report highlighting the growing adoption of Vape Detection Technologies (VDTs) in Minnesota’s school systems. The report illustrated the scope and prevalence of vape detection in public schools in the state of Minnesota with a focus on the privacy and education equity implications of these systems.
In response to the 2024 report, the co-authors of this study, a group of independent researchers and journalists, came together with TCIA and NOTICE to examine marketing claims made by VDT manufacturers — in particular HALO, a leading manufacturer providing “smart” VDTs used in Minnesota schools — about the rationale for and effectiveness of VDTs.
Methods
The co-authors were able to access limited data from schools using HALO systems obtained through a public records request to the Minneapolis public school system. In addition to that limited data, we drew on: expert analysis of validity of claims made by HALO with reference to air quality standards; research into the leading edge of experimental AI-driven sensing and analysis technologies; and social media data shared publicly by students. The NOTICE Coalition also conducted interviews with key community stakeholders about the dynamics of VDTs in school systems. Summaries of these interviews informed this report.
We were unable to find publicly available evidence analyzing or demonstrating effectiveness of VDT systems outside of the types of information gathered here. This is troubling: as use of these technologies expands, there is a lack of independent third-party data or analysis of VDT surveillance, and of information on how this data has been applied by schools in student discipline or how it might impact educational outcomes. Beyond the lack of independent data effectiveness, the use of VDT systems opens up questions regarding privacy rights, responsible stewardship of student data, the safety of the cultural and learning environment, and the relationship between educators and students.
Acknowledgements
The authors of this report would like to express our gratitude to all those who have contributed to this project. Special thanks to Ed Vogel; and to the Twin Cities Innovation Alliance, NOTICE Coalition, ETICAS Foundation, Civic Tech Field Guide, the Georgetown Center on Privacy and Technology, and the Just Tech Fellowship Program for providing material resources necessary to make this report possible. We also extend our heartfelt gratitude to all of the students, educators, administrators, and relentless community advocates in the Twin Cities and beyond who have courageously shared their stories and struggles for the futures we all deserve.