Research Data Policy

Overview

Audio and video data is acquired by Basis Research Institute, as part of our research on animal behavior, especially focused on rats and birds. This document outlines policies and practices related to the data, how it is collected, and how it may be used.

A) Purpose

The purpose of acquiring this data is to understand the behavior of non-human animals. Basis Research Institute studies animal cognition and social behavior as part of our Collaborative Intelligent Systems project. We are interested in both basic science questions (e.g., how do different species of birds communicate when they forage for food together in groups?), as well as applied questions (e.g., how to minimize human-rat conflicts, and which pest-control interventions are most effective, considering both human and animal welfare).

B) Authorized use

The authorized use of the data is the analysis of non-human animal behavior. Relevant environmental data (e.g., sound of rain, surface temperature, etc.) may also be analyzed. Prior to data use, an automated redaction step must be completed to remove incidentally recorded identifiable and/or intelligible human signals before analysis. Aggregate information about the presence of human signals (i.e., timestamps when speech was detected, with speech content and individual voices redacted) may be analyzed as it may influence non-human animal behavior. Prohibited uses of the data include analysis of the identifiable and/or intelligible contents of human signals.

C) Data collection

The data collected is audio and/or video data, including in spectral bands outside the range of human speech/hearing and vision. Specifically, ultrasonic microphones collect audio data sampled at 192 to 384 kHz, which allows for maximum detected frequencies of 96 to 192 kHz (human speech energy is concentrated near 4-7 kHz, and maximum hearing frequency is approximately 20 kHz). Thermal cameras collect video data that include a wide range of infrared wavelengths, calibrated to indicate the temperature of each pixel.

D) Data access

Data will be handled by trained research personnel at Basis Research Institute. In cases where there is a possibility that incidental human signals were recorded, data will be processed via a redaction procedure to remove any incidentally recorded human signals that may be identifiable and/or intelligible. We use “redaction” and “human signal removal” interchangeably in this document to refer to the process of removing identifiable/intelligible human signals from our data. Redaction will occur within a month of recording, and will be done automatically to the extent possible. Portions of the redacted data may be shared with collaborators and/or the public, including through analysis reports. Anyone using the data must consent that they will use the data for its authorized use only (animal behavior analysis).

E) Data protection

Data is encrypted on any unattended recording device, in case the device may be stolen, to prevent use of the data by unauthorized people. Encryption keys are located on secure servers, only accessible by trained research personnel at Basis Research Institute. Once recorded, data is transferred to secure servers by trained research personnel at Basis Research Institute, and is processed to remove any incidentally recorded human signals that may be identifiable and/or intelligible. In order to minimize the chances of encountering human signals, researchers process data prior to subsequent analysis, and perform the human-signal removal automatically to the extent possible (i.e., in the unlikely event human signals leaked through the redaction procedure, researchers would manually remove them from the dataset). Access to data servers is overseen by Basis Research Institute, and associated with each researcher’s Basis email account login credentials. Account access reviews occur periodically (at least annually, and at major staffing changes), access is denied for researchers no longer directly involved in the project.

F) Data retention

Raw data will be retained for up to one month to verify that the redaction procedure successfully removed human signals. We require one month to validate the redaction procedure because, while it will be an automated process that occurs once data is uploaded to our secure servers, different sites with different types of background noises may require different settings to satisfactorily remove incidental human signals while retaining animal signals, and all automated anonymization must be hand-validated by a trained researcher by examining a sample of the data.

Redacted data (no identifiable and/or intelligible human signals) will be retained to support future research. This is a requirement of certain funders of research on animal behavior, including federal funding agencies, to ensure reproducibility and lasting impact of the work.

G) Public access

Raw data cannot be accessed by the public. Select redacted data (no identifiable and/or intelligible human signals) may be shared with the public, with the goal of furthering scientific and community understanding of urban animal behavior.

Basis Research Institute shares reports generated from redacted data with municipal and community partners. Specifically, we share written and verbal reports on rodent activity in sites relevant to them with Boston Rat Action Plan, NYC Parks Greenthumb, NYC Department of Health, and local campus and integrated pest management partners. These entities may use the data to further their goals consistent with the data use policies stated herein (e.g., minimizing rat-human conflict, and enhancing the science and practice of integrated pest management). Sharing of our written and verbal reports by recipients with other City agencies and non-City entities is allowable with the explicit permission of Basis Research Institute and a justification from the recipient explaining how sharing is consistent with the data use policies stated herein. Sharing of our written and verbal reports by recipients with third-parties without both a justification and the prior permission of Basis Research Institute is prohibited. Recipients are further prohibited from attempting to use the data we collect to track, analyze, or otherwise surveil human activity in any way.

I) Training

Only trained researchers at Basis Research Institute have access to keys for data decryption. Before analyzing the data, researchers run automatic scripts to remove any incidentally recorded identifiable and/or intelligible human signals. Training consists of review of this data policy, instruction of how to manually remove incidental human signals if necessary, practices for data security hygiene, and practices for communicating about data with community stakeholders. These training steps are organized in a training document that all trainees have access to.

J) Oversight

This project is overseen by Dr. Emily Mackevicius. She keeps internal records of who received training, who has access to the data, the timeline and performance of redaction procedures, and what analysis has been carried out. The project is reviewed annually by Basis and external project advisors, and independent external academic colleagues are consulted for scientific and data use advice. If anyone has suspicions of violations of the data use policy, they can be reported to contact@basis.ai, and Basis will conduct a formal review to determine whether data should be deleted, and/or whether individuals should lose access to the data. Any individuals found to have knowingly or unknowingly violated this data use policy will lose access to raw data, with the possibility of repeating training to regain access at the discretion of Dr. Mackevicius.

K) Legal Authority

This policy is intended to comply with applicable federal, state, and local laws governing audio recording, video recording, surveillance technology, privacy, data retention, and disclosure. The project is designed for the study of non-human animal behavior and is not intended or authorized for the surveillance, identification, tracking, or analysis of people.

The primary local authority for Boston data collection is the City of Bostonʼs Community Control Over Surveillance ordinance, Boston Municipal Code §§ 16-63.1 et seq., including the requirement that a Surveillance Use Policy specify the legal authority controlling the collection, retention, processing, release, or disclosure of surveillance data. For deployments or partnerships in other municipalities, relevant local authorities may also include the New York City Public Oversight of Surveillance Technology Act, New York City Administrative Code provisions governing NYPD surveillance technology impact and use policies, and the Cambridge Surveillance Technology Ordinance, Cambridge Municipal Code Chapter 2.128. Other relevant legal authorities include the federal Wiretap Act, 18 U.S.C. §§ 2510-2523; Massachusetts General Laws chapter 272, § 99; New York Penal Law article 250; Katz v. United States, 389 U.S. 347 (1967); and Commonwealth v. Hyde, 434 Mass. 594 (2001). Together, these authorities inform this policyʼs restrictions on recording, retaining, analyzing, or disclosing identifiable and/or intelligible human signals, and the projectʼs use of public-facing signage to provide notice of recording activity, identify the research purpose of the technology, and make clear that the technology is not intended for human surveillance.

L) Child Rights

Special consideration is taken to avoid the possibility of recording children. Sites and recording times will be selected to avoid recording within earshot or line of sight of locations where children are present. Before recording rats near any sensitive sites (schools, daycares, playgrounds, youth centers, school bus stops, etc.), we will coordinate with school and/or parks administration.

M) Site Selection and signage

Recording sites are selected to record animal behavior, while minimizing the chances of incidentally recording humans. This is done by a variety of methods, including recording primarily at night, preferring sites that are locked to people, recording at sufficient distance from sensitive locations such as private residences or schools, and posting clearly visible signage near recording devices alerting the public to their presence.

N) Training of computational models

We are researching animal cognition, which involves computational analysis and modeling of animal behavior data in a range of urban social contexts. Those interested can find links to our publications on our project page. Prior to computational analysis, all data will undergo a redaction procedure to remove incidental human signals, and the absence of these signals will be validated by a trained researcher.

Contact

If youʼre interested in this research, and have any questions, concerns, or appreciation, please contact us at rats@basis.ai.