ai-distress-detection

AI Distress Detection

AI distress detection refers to systems that try to infer that a person is in danger from sensor data and trigger an alert automatically — without the user having to press anything. Audio-based variants run scream, gunshot, or “distress” classifiers built from MFCC and pitch features and SVM/CNN models; wearable variants (e.g. smartwatch systems such as Suraksha) fuse accelerometer, gyroscope, heart-rate, and GPS signals to infer assault or a fall. The appeal is obvious: in an attack a victim may be unable to reach a Panic Button, so automation promises help when manual triggering fails.

The critiques are substantial. Reliability is bounded by training data: environmental noise misclassifies vocal cues, and acoustic detectors notoriously confuse fireworks and car backfires for gunshots — EFF notes that over 99% of one vendor’s alerts produced no police action, and that false positives can escalate a police response dangerously. Dataset bias means systems can perform worse on certain demographics, cultures, or vocal patterns, a fairness problem that maps onto Racial Capitalism critiques of who gets surveilled and who gets served. Always-on listening also raises eavesdropping-law and consent issues, putting the technology in direct Privacy and Safety tension and feeding Techno-Solutionism: a confident-sounding model substituted for harder structural work. As a capability it sits inside Personal Safety Apps and overlaps with Safety Wearables, and its accuracy-versus-autonomy trade-offs are exactly the kind of claim a project like The Safest Woman Alive interrogates.

In this vault

Sources

https://pmc.ncbi.nlm.nih.gov/articles/PMC12251837/
https://www.eff.org/deeplinks/2025/10/flocks-gunshot-detection-microphones-will-start-listening-human-voices
https://wjarr.com/sites/default/files/fulltext_pdf/WJARR-2025-3887.pdf

Tags: #concept #ai #detection

Backlinks

risk-assessment ×2
concept-map

Linked Pages

privacy-and-safety

Privacy and Safety

Privacy and safety are often treated as the same goal in anti-violence technology, but for a Victim-Survivor they pull in opposite directions and must be balanced deliberately. The data that makes a safety tool useful — real-time Location Sharing, cloud backups, message logs kept as Evidence Documentation — is exactly the data an abuser will try to reach if they retain access to a device, account, or family plan. NNEDV’s Safety Net Project frames technology safety as protecting “safety, privacy, accessibility, and civil rights” together, and stresses that the burden should fall on holding perpetrators accountable, not solely on survivors restricting themselves. Yet practical safety planning still requires hard privacy choices: which device is trustworthy, which accounts are compromised, and what should never be stored where an abuser can find it.

This is why every safety app sits in tension here. A design that collects rich location history to “build evidence” also builds a honeypot that Stalkerware or a coercive partner can exploit, and the Datafication of MeToo critique shows how such collection can become extraction rather than protection. Privacy-protective alternatives respond with data minimisation, on-device storage, encryption, and Discreet Access, and safety guidance cautions that even removing monitoring can escalate danger if an abuser feels they’ve “lost control.” Getting this balance right is the central problem of Survivor-Centered Design and Trauma-Informed Design.

In this vault

Tension withPersonal Safety Apps
Threatened byStalkerware
Related toDiscreet Access
Related toSurvivor-Centered Design

Sources

the-safest-woman-alive

“The Safest Woman Alive”

Reference: Linnea Öhlund & Angelika Strohmayer (2025). “‘The Safest Woman Alive’: A Reflection on Interpersonal Safety Technologies for Gendered Violence Protection.” CHI EA ’25: Extended Abstracts of the CHI Conference on Human Factors in Computing Systems, Article 651, pp. 1–11. DOI 10.1145/3706599.3716240. (alt.chi)

Summary

A critical, feminist reflection on how HCI designs against gendered violence (GV). The paper opens with a satirical prologue — a woman dressing for a walk who straps on a body sensor, a “smart girl security system” belt, smart-foot devices, multiple panic wristbands, a tazer-watch, a filming ring, pepper-spray GPS sticks, and a dozen “safe route” / panic-button / evidence-recording apps — each one drawn from a real published HCI safety project. The absurd accumulation makes the argument: HCI over-engages with interpersonal safety technologies, loading the burden of protection onto potential victim-survivors and treating a socio-political problem as a personal, gadget-shaped one.

To reframe the field, the authors apply Patricia Hill Collins’ Four Domains of Power (structural, disciplinary, hegemonic, interpersonal) and show that almost every device in their prologue addresses only the interpersonal domain — and mostly the hyper-visible case of stranger street harassment — when the most common GV is intimate-partner and hidden. They warn that such designs are techno-solutionist and can be victim-blaming, demand normative bodies and “correct” use, and rest on unreliable tech (facial/emotion recognition). They call for HCI to design across all four domains — including for perpetrators and structural change — in allyship with support services, and offer a reflective framework spanning a trajectory “from gadgets to critique.”

Key ideas

The burden of GV prevention is wrongly placed on women and gender-nonconforming people, not the institutions and perpetrators upholding violence.
The Four Domains of Power is a usable reflective tool: map which level your intervention actually addresses.
Most safety tech targets the visible interpersonal domain (stranger danger), missing the structural, disciplinary, and hegemonic domains.
“Safe route” routing encodes whose safety counts (ACLU backlash to Microsoft’s version); emotion/face recognition is unreliable and non-inclusive.
Reporting/mapping platforms (Harassmap, Hollaback / Right To Be, Protibadi) gain power from activist use, not data collection alone.
Uses “gendered violence” rather than “violence against women and girls” to include queer and gender-nonconforming people.

Connections

CritiquesPersonal Safety Apps
CritiquesTechno-Solutionism
IntroducesFour Domains of Power
AddressesGender-Based Violence
DiscussesReporting Apps
Informed byFeminist HCI
Related toTrauma-Informed Design
Related toDatafication of MeToo
- Companion critique: both read anti-violence tech through power and political economy.

Sources

https://dl.acm.org/doi/full/10.1145/3706599.3716240

safety-wearables

Safety Wearables

Safety wearables are pendants, bracelets, clips, and “smart jewelry” that put a discreet emergency trigger on the body rather than buried in a phone. invisaWear’s charm, pressed twice, texts up to five emergency contacts a GPS location and can dial 911 directly; the Flare bracelet pairs to a phone, shares location, runs a check-in safety timer, and can fake an incoming phone call with a pre-recorded message to manufacture an exit; Wearsafe’s tag streams location and audio to a contact network on a subscription model. Most are paired Bluetooth tags that lean on a companion app and so are essentially a hardware Panic Button for Personal Safety Apps, typically backed by Location Sharing and a Trusted Network.

The critiques are practical and political. False alarms are common enough that devices ship PIN-based “mark me safe” cancellation, and reviewers report that wearing one can manufacture anxiety where there was none — a felt cost rarely counted in marketing. Cost and subscription fees gate access, raising equity questions that connect to Racial Capitalism and to who can afford to buy safety as a product. Most fundamentally, the model places responsibility on the individual at risk rather than on perpetrators or institutions — a Techno-Solutionism pattern that projects like The Safest Woman Alive put under scrutiny — and the same location and audio capabilities that protect can be turned into Location Tracking Abuse, so they must be weighed against Privacy and Safety and the realities of Coercive Control.

In this vault

Related toPersonal Safety Apps
ImplementsPanic Button
Critiqued byThe Safest Woman Alive
Tension withPrivacy and Safety

Sources

https://invisawear.com/
https://gizmodo.com/wearing-this-ugly-smart-necklace-helped-me-deal-with-my-1834118959
https://blog.squaretrade.com/high-tech-personal-safety-devices/

Tags: #concept #hardware #wearable

personal-safety-apps

Personal Safety Apps

Personal safety apps are mobile applications that help a user summon help, share their whereabouts, or deter harm. Most combine a Panic Button, Location Sharing, and a Trusted Network, and many connect to Crisis Hotlines and Resources. They span two broad postures: alarm-and-summon tools aimed at acute danger, and network-and-intervene tools rooted in Bystander Intervention that aim to prevent harm before it escalates.

The central design tension is Privacy and Safety: the same capabilities that protect (continuous location, contact lists, always-on microphones) can be turned against a user — see Location Tracking Abuse and Stalkerware — so survivor-facing tools must be weighed with Trauma-Informed Design in mind.

In this vault

Member apps assert instance_of::[[Personal Safety Apps]]; see backlinks.

techno-solutionism

Techno-Solutionism

Techno-solutionism is the assumption that complex social and political problems can be fixed with technology. In anti-violence tech it shows up as the belief that the right gadget or app can “prevent sexual assault” or “ensure the safety of women” — a framing critiqued at length in The Safest Woman Alive and Datafication of MeToo. The danger is twofold: it shifts the burden of prevention onto victim-survivors (who must buy, learn, and correctly use devices), and it draws attention and funding away from structural change addressed by the other Four Domains of Power.

This is the central tension running through the vault: tools can genuinely help in a crisis, but should be weighed against Trauma-Informed Design, feminist critique, and the limits of what design alone can do.

In this vault

Critiqued byFour Domains of Power
Tension withPersonal Safety Apps

racial-capitalism

Racial Capitalism

Racial capitalism is the argument, developed by Cedric J. Robinson in Black Marxism (1983), that capitalism did not negate the racialism of the feudal European order but grew within it, so that capital accumulation continually generates and reproduces racial hierarchies. Robinson took the phrase from analysts of apartheid South Africa and generalised it: race and class are inextricable, and “racial capitalism” names modern capitalism as such, not a special variant of it. Robin D. G. Kelley’s gloss is that capitalism was “racial capitalism” from the beginning — its tendency was always to differentiate and exploit through racial regimes, not to homogenise.

In this vault the concept supplies the critical lens for reading anti-violence technology against the grain. Henne, Shelby and Harb’s study of Reporting Apps argues that these tools reproduce racial capitalism: they extract data from users, build property relations on data ownership and accumulation, commodify “diversity and inclusion,” and sustain ties to the Criminal Legal System that carry racialised risks. The lesson is that the seemingly neutral act of “building evidence” or “collecting data” is never neutral — it shapes whose stories become legible and who profits. This frames Techno-Solutionism as not merely naive but materially interested.

In this vault

Sources

Tags: #concept #racial-capitalism #critical

panic-button

Panic Button

A panic button (or SOS trigger) is the central primitive of most Personal Safety Apps: a single fast action — a screen tap, a hardware-button sequence, a voice command, or a wearable press — that raises an alarm and usually shares the user’s whereabouts with a Trusted Network or emergency dispatch. Design tensions include speed versus accidental triggering, and visibility versus the need for discreet activation under threat.

In this vault

Part ofPersonal Safety Apps
Paired withLocation Sharing