Recordkeeping in Voice-based Remote Community Engagement

Driven by pragmatic, cost-related, and environmental factors, voice-based remote community engagement tools (such as Interactive Voice Response) are emerging as a key modality for engaging marginalized communities. These voice-based digital solutions offer new opportunities for distributed community engagement and empowerment, and the ability to capture, store, and access a wide range of different records (i.e., recordings, interactions and contextual metadata) associated with community engagements. This potential for large scale, distributed community record collection necessitates an understanding of inclusive and effective recordkeeping approaches for appraisal, documentation, preservation, and accessibility of different types of records (such as audio recordings, transcripts, reports, and observatory notes) related to voice-based community engagements. Through qualitative analysis of stakeholder focus group discussions with domestic workers (as marginalized community members) and NGOs working in the sector, we present valuable insights and recommendations for the development of recordkeeping approaches tailored to voice-based remote community engagement records.


INTRODUCTION
An emerging practice in international development involves leveraging digital technologies for community engagement initiatives to facilitate knowledge sharing, education, training and informed decision-making [88].An increasing proportion of community engagements, such as skill development initiatives, communitybuilding discussions, and training and awareness programs between non-government organizations (NGOs) and their target rural communities, are held remotely [18,99].This shift is often driven by practicality, cost considerations, a desire to minimize environmental impact, and public health compliance [15,78].While parts of the world have been able to use internet-based technologies such as online video chat to facilitate remote engagements, many communities still rely exclusively on voice calls through traditional telephone systems for remote communication [81,82].This reliance can be due to a wide array of potential socio-technical impediments, including: limited access to smart devices or mobile internet coverage [101]; cultural barriers to phone access and use [95]; and low digital literacy [5,53].In response, a number of initiatives have attempted to respond to these challenges by combining traditional telephony with digital infrastructures: often utilizing Interactive Voice Response (IVR) interfaces to support users in accessing information through button presses over phone calls [67,74,108].
While such platforms have unlocked new opportunities for distributed community engagement and empowerment through these voice-based 1 (i.e., telephony) technologies, they also ofer opportunities for these engagement interactions (including corresponding contextual metadata) to be captured, stored, and made accessible as authentic and reliable records2 of activities during engagements.Community recordkeeping, where the appraisal, documentation, preservation and disposal of records are undertaken alongside stakeholder communities in a participatory process [17,29,60], ofers the potential to empower communities through opportunities to: ensure that records fairly and accurately represent their lived reality; use such records as citable evidence towards their own agendas; and enforce transparency and accountability upon the organisations who engage with them [12].However, the application of such community-oriented participatory recordkeeping approaches is yet to be explored, particularly in the context of these voice-based platforms.
Numerous concerns need to be addressed when developing and implementing processes to support participatory recordkeeping practices via these voice-based platforms.For example, how could such platforms support stakeholders' participation in recordkeeping of their own engagement records, without relying upon them having access to the internet or higher levels of literacy?Similarly, how can consensus be achieved by stakeholder communities and NGOs as to what records should be preserved, keeping community members in-theloop?Exploration of these aspects of participatory recordkeeping holds particular signifcance for marginalized communities, who usually fnd themselves in passive roles subject to organizational policies and interests [19,88].
To this end, through qualitative analysis of a series of stakeholder focus group discussions with domestic workers (as marginalized community members) and NGOs working in the sector, this paper seeks to investigate: i) the requirements and apprehensions of participating community members, contrasting them with those of the staf from engaging NGOs, that would need to be accounted for within development of community-oriented participatory recordkeeping processes; ii) the existing recordkeeping cultures within the NGOs and the engaged communities, along with their perceived value (if any) in recordkeeping particularly of their voice-based engagement records (including recordings, transcripts, and observatory notes taken during engagement).This paper contributes insights from these stakeholder engagements, and recommendations for the development of community-oriented participatory recordkeeping processes for potential use with voice-based engagement platforms.

RELATED WORK 2.1 Recordkeeping and HCI for Development
Research in the domain of Human-Computer Interaction for Development (HCI4D) has a signifcant focus on the design and deployment of technologies aimed at enhancing the socio-economic conditions of distributed marginalized communities [103].In this context, it becomes imperative for community-centred organizations (including local and international NGOs and government organizations) to have reliable recordkeeping systems that can efectively collect, access, and disseminate records related to their community engagement activities to make informed decisions, track progress, and demonstrate the impact of their interventions on the development of these distributed communities [9].For instance, to carry out crucial economic and social endeavors by these organizations, a well-structured recordkeeping system is vital to support timely access to necessary information [92].According to Thomassen (2015, p. 84) [26], recordkeeping, also referred to as archival science, investigates the cultural signifcance of archival materials and seeks to answer questions such as "why, how, and under what circumstances human beings create, keep, change, preserve, or destroy records, and what meanings they may individually or jointly attribute to records and to their recordkeeping and archival operations".However, conventional recordkeeping, as outlined by ISO 15489, tends to concentrate predominantly on organizational structures and needs, often sidelining user or community perspectives [6].
Conversely, personal information management (PIM), a related feld developed in the 1980s, focuses on efcient digital information management, addressing the needs and concerns of individual customers regarding the management of an abundance of individual digital records [8,47,49,57].Related studies on PIM have explored recordkeeping practices in the healthcare domain: including patients' record collection by family caregivers [33] and review of patient's health records [13,44]; and recreational activities: including gourmet cooks' information and documentation practices [41,42], wedding planning calendars [62], and household tracking through planners and calendars [63,64].Individual tools such as smart journals, a tool for creating and curating digital records of individual's personal life, have been used to explore how individuals manage and value digital historical records of their personal life [27], and a systematic review has investigated how mobile browsing and usage assists individual users in fnding records, as well as how users keep those records in their mobile devices [1].These studies addressed challenges stemming from the abundance of records and viewed individuals as potential clients of recordkeeping tools and services, including digital technologies [47,56].However, all these studies, emphasizing individualistic recordkeeping in technology and information-rich contexts, sharply contrast with contexts of marginalized communities who have limited access to information and technologies, highlighting the need for community-oriented participatory recordkeeping practices.
The term "participatory recordkeeping" has emerged from development intervention studies in western urban and NGO contexts in developed countries, where its primary objectives revolve around community empowerment and democratizing knowledge production [85].Participatory recordkeeping particularly emphasizes the need for collaborative decision-making and recognition of local recordkeeping practices to ensure sustainable and culturally sensitive recordkeeping processes [59,85].One example, CivilServant is an experimental infrastructure facilitating online communities and their moderators in policy evaluation, fndings replication and data governance of Reddit, addressing ethical and participatory concerns pertinent to recordkeeping practices within online platform interventions [59].Similarly, Citizense Makers is an online platform designed to facilitate community-based recordkeeping by enabling users to engage in collaborative sensemaking through reciprocal sharing, annotation, and discussion of their records of self-tracking data.The platform aided community members to collectively interpret and manage their records [79].While these limited examples of participatory recordkeeping have addressed several aspects of recordkeeping in the context of online communities (including appraisals and sharing of records), recordkeeping practices and implications in the context of community engagement activities (such as life-skill training) involving NGOs and marginalized community members are yet to be explored in the literature [35].
A limited number of HCI4D studies has addressed individual elements of recordkeeping including: information rights [87]; Artifcial Intelligence and archives [14,20]; personal digital collections [106]; and archival data modeling [52].These works however offer limited insights for participatory recordkeeping as a wholistic activity.Consequently, in the domain of HCI4D, which prioritizes community engagement and empowerment, there exists a signifcant gap in acknowledging and adequately addressing the needs and development goals of distributed marginalized communities, particularly within the context of community-NGO participatory recordkeeping practices.

Existing Privacy & Records Access Practices
The feld of Usable security (USEC) research can provide insights for participatory recordkeeping: fostering inclusivity, ownership, and ethical engagement by ensuring accessible and culturally sensitive security features and user interfaces [31,32].USEC refers to the seamless integration of privacy and security features into technological solutions, ensuring users can easily comprehend and apply these features [73,89].It focuses on the interactions of cybersecurity and usability; encompassing usability challenges related to authentication, encryption, social engineering, security dialogs, and privacy [58,105].For instance, USEC research has evolved with human-centered approaches in response to the users grappling with managing numerous password-protected accounts [93], followed by the adoption of diverse contemporary authentication methods including biometrics [32,90].Drawing parallels to USEC research, which has sought to cover human factors in the form of social engineering [51], secure browsing [4], and online privacy [37], there is a similar gap in literature addressing recordkeeping challenges for marginalized communities which need to be addressed.This gap emphasizes the necessity for HCI researchers (as USEC researchers did before us) to similarly advance approaches to suit the unique requirements of communities in terms of participatory recordkeeping [68].
Another area of interest in this domain, that is relevant for our discussion, is how communities (and in particular, marginalized communities) access state records.This is a challenge globally for the feld of recordkeeping, and much needs to be done to shift theoretical rhetoric to practical, community-oriented recordkeeping applications.For instance, the experience with indigenous groups in Australia refects a need for nuanced approaches that acknowledge agonizing historical contexts and value community perspectives in record evaluation [21,71].Similarly, the issue of 'empty archives' resulting from constitutional mandates3 in South Africa advocates for development of inclusive archival holdings, emphasizing the importance of recordkeeping ecosystems that represent the diversity of communities nationwide [69].Besides, one comparative study regarding the accessibility of the state records of Zimbabwe and South Africa reveals a signifcant under-utilization of oral sources (i.e., histories of marginalized people), prompting recommendations for adopting information and communication technologies (ICT) in ensuring inclusive and equitable accessibility of state records (including oral sources) [10,69].
While most related studies highlight global concerns regarding collection and accessibility of state records, the intersection of ICT for development (ICT4D) and archival science unveils a critical gap in preserving records from projects aimed at enhancing socio-economic status of marginalized communities in developing countries [35].The creation of analog records by community participants in a Bangladesh-based ICT4D project exemplifes communitydriven eforts to overcome this gap, contributing valuable insights to interdisciplinary research between ICT4D and archival science [35].Additionally, the distinct contexts of state records and (NGOled) project-based community engagement records underscore the imperative for a more comprehensive and community-oriented approach to participatory recordkeeping [38,70], particularly within the latter context, which remains underexplored in the literature yet.

Voice-based Community Engagement
Recognizing the crucial role of incorporating community dialogue in international development, as well as acknowledging the obstacles faced by communities lacking internet connectivity and literacy, voice-based media has gained considerable attention as a tool to support distributed community engagements in the global south [88,100].IVR services evolved as a primary means of asynchronous voice-based engagement to provide benefcial services to underserved communities [24,28,81,95].Specifcally, IVR-based services have been used to create distributed social networks, information accumulation, and dissemination platforms that encompass multifarious application domains such as education, health, agriculture, politics, fnance, journalism, and entertainment [55,74,76,83,84,94].For instance, Avaaj Otalo [74] provided farmers with a IVR-based platform to communicate their concerns with other agricultural community members.Avaaj Otalo observed that farmers exhibited greater interest in peer-to-peer learning, requiring creation and sharing of audio recordings.Similarly, CGNet Swara [67] ofered an IVR-based citizen journalism platform, where community members could report and listen to verifed stories (i.e., access and sharing of recordings) of local interest over the phone and Mobile Vaani [65] ofered an IVR platform for social responsibility and awareness campaigns, where people could record a message, listen to own and others' recordings, and share recordings with others along with commenting on others' recordings; thus, encompassing various facets of participatory recordkeeping.Ila Dhageyso [39], another IVR-based platform created to assess Somaliland's media climate and geopolitical conditions, enabled community members to listen to contents published by the Presidential Ministry and record their own thoughts or reactions to them.Considering the sensitivity due to political context, these audio recordings were reviewed carefully and posted on YouTube following moderation; thus, highlighting the importance of moderation of records.Besides, the transcripts of the recordings were annotated to identify positive and negative sentiments of the community, as most recordings were found to be emotion-rich contents.However, this process of moderation, preservation, and sharing of the records obtained from these asynchronous IVR-based platforms remains unexplored in the literature and specifc to implementations.
However, being asynchronous in nature, IVR-based engagements could not promptly address user-specifc inquiries, thus lacking real-time engagements [74,83].These limitations were largely addressed by synchronous voice-based engagements [48,95], where community members can dial in with their available phones to listen to a live engagement session, ask questions, and interact with a host controlling the show from a smartphone or web application.For instance, Sehat Ki Vaani [48] allowed social workers to invite healthcare providers (HCPs) as guest speakers on a telephony broadcast (i.e., synchronous session), similar to a radio show, where listeners could join the call and ask questions.Similarly, Citizen Radio (CR) [108] and Sangoshthi [95] also explored synchronous community engagement using telephony leveraging diferent roles (e.g., host, listeners, and HCPs as guests), selection of topics during live session, and generation of relevant records from the live session (e.g., session recordings, transcripts, interaction records from system logs).Another synchronous platform, LEAP [107], facilitated peer-learning of community health workers through playing three recorded audio clips on relevant topics, each followed by a live group discussion.Thus, synchronous community engagement platforms facilitated creation of a wide range of engagement records during live sessions, distinct from those of asynchronous engagement platforms [107].In addition, Sehat Ki Vaani and Sangoshthi broadcast a recorded trailer to advertise the upcoming session, unlocking the potential for (reusable) content creation (e.g., session summaries, podcasts) with the recordings of the synchronous engagements.However, similar to asynchronous platforms, they did not address the critical issue of recordkeeping of these informationrich and potentially sensitive records.
The implementation of both asynchronous and synchronous voice-based activities highlights the signifcance of retrieving and sharing recordings for subsequent reference [82,108].This capability is especially benefcial for individuals with low literacy, who may encounter difculties taking written notes and require frequent access to identical information.However, the literature is yet to explore how the voice-based engagement records (including recordings, transcripts, facilitator notes, and interaction metadata) obtained from these voice-based distributed community engagements can be moderated, preserved, utilized, and shared for multifarious purposes including community empowerment through sustainable knowledge gain, organizational evidence, and reporting.

METHODOLOGY 3.1 Field Sites and Participant Details
The research study took place in Dhaka, capital city of Bangladesh, a South-Asian country.The research was conducted in collaboration with 'Nari Maitree (meaning, Women's Solidarity)' 4 , an NGO that works with one of the most marginalized and under-represented communities-domestic workers based in Dhaka to assist them in securing jobs and alleviating literacy barriers.Nari Maitree primarily focuses on skill-building initiatives and community training and discussions covering topics like labor rights, health, climate change, workplace conditions, and social integration.Voice-call sessions can serve as a platform to synchronously or asynchronously run their community-building discussions around various topics, and share collective experiences, challenges, and aspirations during these discussions as well as contributing to the co-creation of knowledge [48,108].In addition to remote engagements to connect hard-to-reach communities and during any restrictions regarding in-person engagements, Nari Maitree deems voice calls particularly useful for follow-up discussions upon any in-person engagements.
The records generated during these engagements capture crucial insights and collective knowledge of communities, serving as a repository for reusable shared information and a reference for future projects and employment opportunities.Hence, the motivation to access and amend these records is closely tied to the tangible impact on the community participants' lives, ensuring accurate representation and preservation of their perspectives and knowledge.For instance, amendments often relate to corrections of personal information, updates on employment status, or documentation of skill-acquisition milestones achieved through the NGO's programs; making the community participants open to potential future opportunities.
During COVID-19, most of Nari Maitree's community engagement activities were conducted remotely over the phone (e.g., group calls).There are 10.5 million domestic workers in Bangladesh, and 90% of them are female workers [72].Domestic workers, mostly living below the poverty line, come to Dhaka from diferent parts of Bangladesh in search of jobs.They mainly reside in slum areas to aford the living costs in Dhaka, have poor living conditions, and are deprived of labor rights such as no formal employment contract and limited paid leave [97].Nari Maitree's ofcial space served as the venue for the research activities, considering the convenience of all participants in commuting.
The study involved 18 stakeholder participants in total -nine domestic workers and nine NGO staf (who host and facilitate diferent engagement activities with domestic workers).The participants are all residents of Dhaka, and native Bengali speakers.The domestic worker participants are originally from diferent parts of Bangladesh, including Southern (Barisal, Bhola districts) and Northern (Mymensingh district) parts of Bangladesh.The literacy level was mixed: some could fairly read and write, some could just write their names and signatures, and some were illiterate.The NGO staf participants (we will call them facilitators onward) were from two diferent NGOs: Oxfam 5 and Nari Maitree, and are all literate persons.The authors have collaborated closely with both NGOs previously.Despite no personal acquaintance with any participants, the authors' prior collaboration with both NGOs endeavors to ensure ethical research practices and a deep understanding of socio-cultural dynamics.Besides, the inclusion of facilitators from two diferent NGOs aims to encompass a broader spectrum of NGO perspectives, contributing to a more holistic understanding of the collaborative dynamics in engaging with domestic workers.Table 1 shows the demographic information of 18 stakeholder participants.The research activities and materials were reviewed and approved by the ethics board of the authors' host institution.Nari Maitree recruited all of 18 participants in careful consultation with the authors.As part of the recruitment process, Nari Maitree circulated the research explanatory statement (including the research goals, participants' involvements, data collection and confdentiality policies, and contacts and complaints) and consent form (electronically to facilitators and verbally to domestic workers) within their collaboration networks.The recruited domestic workers and facilitators had extensive job experience in their relevant felds of work.Upon recruitment of the participants by Nari Maitree, the frst author explicitly read aloud the explanatory statement and the consent form in the venue at the onset of research activities to ensure voluntary participation of all participants as well as considering the literacy barrier of domestic workers particularly in understanding engagements in a research context (in contrast to their other regular engagements with facilitators).For facilitators, the consent was taken in written form; for domestic workers, verbal consent was audio recorded.The consent was taken for capturing audio recordings, note taking, and photographs during the research activities, and using these records for research purposes only by the researchers.
The in-venue research activities were conducted in presence of a research team comprising three researchers, who are native Bengali speakers and well-versed in the cultural context of the participants.This research team was responsible for taking consent, explaining activities, taking observatory notes, recording audio, and logistic support during the in-venue research activities.

Research Activities
The research activities included three 2-hour long workshops: frst one with facilitators only, second one with domestic workers only, and third one with both facilitators and domestic workers.The facilitator-only workshop (i.e., Workshop1) concentrated on realizing the existing recordkeeping practices of NGOs and overall perception of facilitators on collection and use of records.The domestic workers-only workshop (i.e., Workshop2) focused on the domestic workers' perceptions and awareness regarding the records generated as part of their engagements with NGO facilitators.The third workshop (Workshop3), combining participants from both stakeholder parties (i.e., mixed workshop), aimed to delve into the disparities in attitudes and opinions between domestic workers and facilitators regarding recordkeeping, as well as identifying points of consensus to inform discussions on recordkeeping of voice-based community engagements (e.g., group calls) including audio recordings, transcripts, and facilitator notes of the engagements, along with implications for voice-based platforms that can bridge their difering perspectives.This focus on voice-based engagements and related recordkeeping implications was motivated by their potential to empower distributed communities (as indicated in Section 2.3), as well as the imperative of cultural sensitivity and inclusivity, particularly concerning marginalized communities like domestic workers.

Workshop1
: Facilitator-only.Workshop1 facilitated focus group discussions (FGDs) encompassing the types of activities commonly run and resulting records collected by the facilitators, their motivations for record collection and usage, record storage and access processes, consent process and communication with domestic workers regarding their record collection and sharing.There were three activities in the workshop.Since all activities were designed to be group activities (i.e., group discussions), the facilitators were divided into three groups of three persons.One researcher from the research team supported the activities in each group.
The frst activity was designed to stimulate thinking about all the diferent kinds of community engagements the facilitators had done.The facilitators wrote their diferent engagement types such as community training, stakeholder meetings, surveys, etc., on postit notes and stuck them on a given poster paper.The second activity targeted collaborative exploration of diferent record types created from their common engagements through informed discussion within each group (Figure 2a).The record types were provided as stickers (Figure 1) to label the post-it notes.
The third activity facilitated a more critical discussion on different record types that the facilitators identifed in the previous activity.In this activity, following "Scenario Co-Creation Cards" [3], the facilitators were provided with a stack of record type cards and a stack of question cards6 regarding records.The stack of record type cards included any new record types identifed in previous activity.
The facilitators took turns drawing one card from each pile, and discussed the picked question for the picked record type in their group.Those two cards were then removed from the piles to avoid repetition.The workshop ended with an inter-group FGD on the third activity, as all three groups were brought together to share their insights.Each group picked their discussed most interesting or challenging question for a particular record type, and facilitators discussed that question altogether.

Workshop2
: Domestic Workers-only.Workshop2 concentrated on domestic workers' perceptions of their engagement records.Particularly, the workshop explored the domestic workers' awareness and understanding of record usage, access control, and The workshop comprised two activities.Considering the literacy barriers of domestic workers, the activities were designed to be completely voice-based.The frst activity was on familiarization with records: stimulating the domestic workers to think about their records being collected and shared and relevant consents.The activity involved playing two versions of an audio recording of a pseudo-conversation between a female patient and a female healthcare customer service agent regarding a doctor's appointment on a sensitive female health issue: one mentioning no consent message for the conversation being recorded and the other explicitly mentioning the consent.Upon playing each version of the recording, some follow-up questions 6 were asked and discussed among the domestic workers within their group.The questions were designed to instigate domestic workers' thoughts on awareness, comfort, and potential behavioural infuences related to records, as well as preservation, sharing, and amendments of records.
After some familiarization and brainstorming with several aspects of recordkeeping in the frst activity, the second activity facilitated in-depth group discussions aimed at gaining deeper insights into domestic workers' perspectives regarding record ownership, utilization, and access control.In this activity, the researcher in each group presented several short fctional scenarios 6 closely related to domestic workers' lived experiences, each scenario covering a distinct record type.Each scenario had some follow-up questions regarding its targeted record type, prompting group discussions among the domestic workers based on their related thoughts and experiences.Overall, the follow-up questions specifcally centred around value, access and sharing, and length of storage of diferent records from domestic workers' perspectives.

3.2.3
Workshop3: Mixed Workshop.While the frst two workshops mainly focused on existing recordkeeping practices and value of records through the separate lens of facilitators and domestic workers respectively, Workshop3, provided both parties with a platform to collectively discuss potential recordkeeping processes in terms of enhanced accessibility, usability, and agency.Specifcally, this mixed workshop sought to fathom where they difer in terms of record types and aspects of recordkeeping and the underlying reasons for these diferences.The aspects of recordkeeping included access, explicit consents, amendments, and preservation of records.The focus was on identifying the common ground and determining the aspects of recordkeeping that hold the highest priority for both parties.
The workshop was split into two activities with 18 participants: 9 domestic workers and 9 facilitators.Similar to Workshop2, the frst activity was scenario-based FGDs in groups (Figure 2b).Three groups were formed combining both parties-3 from each.Each group had one researcher to support the activity.There were four scenarios 6 to discuss in each group: 1) Access to transcripts and amendment process if any discrepancy or inaccuracy is found; 2) Group calls and moderation process of call recordings (e.g., choosing the interesting bits to share); 3) Access to audio recordings, particularly when the primary verifcation identity (the original phone number) is lost; and 4) Scope and process of amendment of audio recordings.Each scenario, prompted with follow-up questions, facilitated in-depth discussions around recordkeeping aspects of voice-based engagement records.
The second activity facilitated refections of participants on the interactions performed in their mixed group, and particularly provided an avenue for participants uncomfortable sharing in the mixed group due to power dynamics [19].Hence, in this activity, participants were divided into two groups: 1) domestic workers only, and 2) facilitators only, each group accompanied by a researcher to initiate the group discussions with specifc questions 6 .Apart from prompting refections on the previous activity, these questions had a particular focus on understanding the design considerations for the voice-based platforms that encompass acceptable recordkeeping practices, since: 1) voice has been identifed as a preferred mode of engagement particularly in the context of culturally sensitive communities (e.g., domestic workers) [2] and 2) both parties indicated their preference and comfort towards voice-based engagements and records.One example question: How would you feel about someone willing to update a recording of a workshop you ran/attended?Would you want to know about it, or have a say?

Data Collection and Analysis
The discussions within each group in all three workshops were audio recorded.Notes were also taken by the supporting researchers during group discussions.The audio recordings of the FGDs were transcribed and translated by the frst author (who is a native Bengali speaker).Refexive thematic analysis was conducted (adopting an inductive approach) [11], as the workshop activities facilitated discussions on participants' lived experiences and perceptions around recordkeeping in context of diferent community engagement activities.The frst two authors independently analysed the transcripts manually to generate the initial codes and themes, followed by subsequent discussion and iterative refnement of codes and themes with the co-authors.Frequent and similar codes were iteratively collated into overarching patterns to fnalize the themes.

FINDINGS 4.1 Record Appraisal and Storage
4.1.1Record Value and Use.The workshops provided a platform for two distinct stakeholder groups to express difering viewpoints regarding the signifcance of record collection and the rationale underpinning its potential use and retention.In Workshop1, a clear disparity emerged among facilitators on perceived value and use of collected records, revealing diverse appraisals of the importance attributed to the records.For certain facilitators, these records embodied tangible proof of their active participation, as articulated by F2: "When we're in any training it's good if our activities are collected and recorded: this will be useful, as we are monitored how we have been participating."(F2, Workshop1) Conversely, other facilitators in the same workshop emphasized the indispensability of these records in safeguarding disseminated knowledge and preserving the trajectory of events during engagements, as F3 highlighted: "If any information is not collected, if we don't preserve it, then it's not remembered in the future." This divergence has practical implications for facilitators who might need to reconstruct engagement details for organizational records.This concern was underscored in Workshop3, where F2 highlighted a practical challenge: "We might need to prepare a report for a meeting, but we haven't got any time to prepare the report in the past two months.Now, after 2 months, I may forget things about that meeting needed for preparing the report." (F2, Workshop3) Workshop1 underscored facilitators' consensus on the value of incorporating diverse record collection methods, including written notes alongside audio recording, to improve organizational recordkeeping processes.Despite acknowledging potential participant discomfort, facilitators recognized the worth of such supplementary metadata (e.g., taking notes alongside recording audio).While acknowledging these drawbacks, F2 contended: "[Taking notes] can create unnecessary stress, and cause the participants to be less expressive." F3 afrmed their importance: "Okay.Still, notes are important." F1 concurred: "You are right, we need them for reporting anyway." Workshop3 consistently highlighted domestic workers' conviction in the value of unhindered access to preserved records.They frmly believed that these records held potential utility in future engagements.F1 posed a pertinent question regarding the necessity of access to records particularly if a follow-up or recall/summary meeting of original meeting is arranged, answering to which DW5 and DW3 reiterated their agreement, emphasizing the necessity of record access irrespective of follow-up meetings: "In a follow-up meeting we can share what decisions were taken or what topics were discussed in a previous meeting that we might have forgotten.Should we be able to listen back to these things later?"(F1, Workshop3) Workshop3 also introduced the concept of using records for verifcation purposes.Participants discussed the feasibility of employing audio recordings to validate the accuracy of the organization's reports and metadata.DW3 introduced this notion, "We'll verify if there is any mismatch."Addressing literacy concerns regarding matching the organizational written reports with audio records, DW3 further explained, "We'll take this to someone who can read." DW1 further confrmed, "We would not recall, but we'll have the recordings to verify back, right?" 4.1.2Length of Storage.Workshop3 unveiled a diverse range of viewpoints on the duration for which engagement records should be retained.Participants shared their thoughts, each ofering unique insights on the complex nature of this issue.The discussion commenced with facilitators explaining how their organizations set specifc timeframes for keeping participant records.F1 suggested, "It should be stored throughout the project's duration."F2 and F6 added details, mentioning that it's "ofcially 6 years" and "7 years, ofcially", respectively.These numbers underlined common practices about how long records are usually kept.Contrastingly, both stakeholder groups expressed a common aspiration for records to be preserved beyond the project's duration.F2 expressed this sentiment by saying, "Yeah, but there should be a system that all recordings are preserved beyond project duration somewhere centrally." This shared aspiration highlighted the understanding that records might have value even after a project ends.
The reasons for records to be kept for a longer time were not the same for facilitators and domestic workers.Facilitators focused on using records for future proof, while domestic workers saw them as opportunities for future involvement.For facilitators, the importance of keeping records lay in their usefulness for future evidence and referencing purposes.On the other hand, domestic workers saw keeping records for longer as an opportunity to stay connected with the NGOs for paid involvement and further knowledge acquisition from the NGOs' future projects, as DW8 shared in Workshop2: "No defnite time, as it will be useful for us and others." DW1 echoed this idea, suggesting that records might open doors for them to take part in future projects.: "We need them beyond the duration of the project, as these documents are the basis of our involvement in future projects and getting our contacts for the next projects.So, it will be better if these documents are kept as long as [the NGO] is there." (DW1, Workshop2) Consequently, domestic workers' interpretation of record deletion exhibited a nuanced aspect.Instead of merely viewing record deletion through the lens of data privacy and security, they attributed it to organizational fnancial motivations, as DW5 expressed during the second activity of Workshop3 (when DW1-DW9 were grouped together): "[Something that surprised me during the group discussion was] how they would delete our recordings, when they are doing big multi-million projects.We feel like we are cutting into their business!"(DW5, Workshop3)

The Value of Physical Transcripts.
A somewhat unexpected topic of discussion was the participants' preference for hard copies of the recording transcripts, printed onto paper.This preference for physical records, indicated in both Workshop2 and Workshop3, was motivated by their perceived advantages over digital copies, a sentiment shared by both domestic workers and facilitators.Participants expressed a strong belief in the reliability of physical records.They deemed physical records more dependable than digital versions due to concerns about potential hardware or software failures associated with digital technologies.DW7 acknowledged this viewpoint: "Actually, anything related to machines can crash, but written things with signatures do not crash." Moreover, the participants identifed another advantage of physical transcripts-suitability for extended storage.They believed that paper transcripts were easy to preserve carefully in physical fles, making them a reliable option for longterm safekeeping, as DW4 afrmed: "Paper transcripts are easy to keep in a physical fle carefully.They can be stored for a long time reliably."F1 concurred, highlighting the resilience of hard copies compared to fragile digital devices: "Right.Machines can be broken or damaged, but hard copies are not damaged easily." The participants also voiced concerns about digital records, emphasizing their susceptibility to unauthorized access and potential breaches.F1 shared concerns about the inadvertent sharing of recordings: "Recordings can be shared with, or somehow reach, the wrong people.", while F2 drew attention to the vulnerability of digital recordings to hacking: "Yeah, that's why hard copies are reliable.Recordings as technology can be hacked."Interestingly, the participants (particularly, domestic workers) underscored the risks associated with digital device theft.They believed that records stored digitally, especially on devices such as mobile phones, were more prone to theft compared to physical records.DW5 succinctly articulated this concern: "Mobiles and recordings can easily get stolen.But papers will not be lost.", and DW4 agreed: "But paper will not be stolen." In light of these considerations, participants unanimously endorsed the preservation of hard copy transcripts over audio recordings for long-term storage.DW6 and F5 both echoed the sentiment that the transcript's durability and descriptive nature made it a more ftting candidate for extended preservation: "The hard copy of the transcript should be kept longer." Recognizing the signifcance of transcripts, participants stressed the importance of accuracy.They expressed a strong desire to cross-check the transcript's contents personally.F7 suggested an option to ensure accuracy: "It can be that both the audio recording and the transcript are given to both parties, so that we can cross-check." This emphasis on accuracy refected the participants' concerns about potential discrepancies.
While participants acknowledged the value of transcripts, they were mindful of an obvious downside of transcripts-literacy limitations.They highlighted the importance of access to audio recordings, and proposed a dual approach: retaining both audio recordings and transcripts to cater to participants with varying literacy levels, as DW5 stated: "It's better to keep both the recording and the transcripts.Those who cannot read can access the recordings, and those who can read can just read the transcript." (DW5, Workshop3) The utility of audio recordings extended beyond transcriptions, as domestic workers saw them as tools to validate reports and metadata accuracy.They recognized their role in validating reports and metadata accuracy, even if personal literacy posed a challenge, retaining a degree of independence from the engaging NGOs.DW6 suggested involving trusted third parties in such cases: "I cannot, but I will have someone who can read." Demonstrating their awareness towards confdentiality, DW3 added: "We can rely on our family members who can read, as this reduces the risk of sensitive information being exposed."

Record Access and Sharing
4.2.1 Participants' Access.It was clear from the discussions in all three workshops that the participants recognized the importance of having access to their engagement records including audio recordings.However, a central concern emerged in the discussions in Workshop3 regarding the secure and accessible means of achieving this, particularly for audio recordings, without imposing signifcant literacy requirements or relying on internet-connected devices and infrastructure.The participants were receptive to the idea of calling into an IVR service to listen to the recordings, leveraging their original session phone numbers for identity verifcation purposes (DW2: "Yes, makes sense, as it was my speech from my number.").Further concerns arose when addressing scenarios where participants lost access to their phone numbers, prompting consideration of potential solutions, with in-person re-authentication suggested by facilitator participants as a viable option.Facilitator participants suggested several potential solutions in this regard, including sharing alternative access information in advance, or in-person re-authentication (e.g, showing photo ID) to access the recordings: "They need to share this information earlier with us: if they lose their number, how or from whom they wish to listen to the recordings." (F1, Workshop3) "The participants usually have a unique ID number-ID and picture.The participant can show their ID, then we can allow them to access the recording." (F7, Workshop3) Options for re-authentication using common secret answers (such as date of birth, or their mother's maiden name) were not deemed to be secure enough, given that the participants wanted to be able to keep some recordings secure even from immediate family, as DW8 expressed: "My husband will be able to guess, even if I put my national ID number!"An agreed alternative was implementing a PIN system, where each recording was protected by a passcode.Despite pragmatic concerns like older participants struggling to remember passcodes, a solution was suggested: writing down the password in participants' preferred language if necessary.This approach demonstrated a balance between security and practicality: "We can have a password for each recording, like 1122 or 1123.So anyone access from any number if the correct password is given.[...] But there are persons who might fnd it difcult to remember the four digits, especially older people." (F8, Workshop3) " They can keep the password written somewhere, in Bengali if needed." (DW9, Workshop3) Participants also demonstrated an inclination towards generating their own recordings, emphasizing the importance of personal agency in contrast to relying solely on having access to organization's records.F5 suggested that creating individual recordings would guarantee continuous access and provide a more authentic representation of their session experiences: "I can also keep the recording by recording on my own phone.Recording myself can be quite diferent to a recording being shared." (F5, Workshop3) In particular, domestic workers expressed apprehension regarding their access to organizational records and deemed the ability to create personal recordings a form of insurance.Consequently, they contended that this autonomy was vital, considering potential future needs to retrieve specifc recording(s).This collective sentiment highlights the participants' strategic approach to recording practices within this context, as demonstrated in the conversation between DW3 and DW1: "What do we do when [the facilitator] has the full recordings, and we want to access them?" (DW3, Workshop3) "We should carefully remember to start recording before talking.Otherwise, we may not get those recordings for further actions." (DW1, Workshop3) Independent recordkeeping was framed as a means to either gather evidence of misconduct when conducted covertly or as a preventive measure to discourage misconduct when conducted openly, as exemplifed by DW3: "Also, when the staf knows that we might be recording as well, he will be careful of doing wrong things with us." Examples of these "wrong things" might include instances of mistreatment, misrepresentation, emotional abuse, exploitation, or any behavior that compromises the social well-being or rights of the domestic workers (e.g., posting sensitive recordings in NGO's ofcial social media page).In essence, the conversations revealed two distinct approaches to maintaining access to recordings, with participants weighing the benefts of organization-provided records against the autonomy of self-created records or transparency in accessing records.This contrast refected not only practical considerations but also the nuanced dynamics between domestic workers and facilitators.

Third
Party Access to Records.The discussions also unveiled several insightful themes, highlighting the dynamics of sensitivity, privacy, and accessibility of records, particularly in terms of third party access.Sensitivity emerged as a paramount concern, especially among participants ensnared in controlling household relationships.The potential for accessible records to fall into the hands of partners raised valid apprehensions about privacy, potentially exacerbating their vulnerability; as in the context of audio recordings, DW2 mentioned, "We may say something that cannot be shared with our sons, daughters or husband, because we say many things in the fow." Regarding being careful about what they speak of and where, DW6 shared, "But talking privately may raise questions to my husband.That can happen too." -the mere act of talking privately over the phone could arouse suspicion and raise questions from their partners.DW7 added, "So I would just say over phone that my husband is at home, I will talk later not now.".This sentiment underscores the delicate balance participants must strike between open communication and the need to navigate their household relationships safely.
Amidst these concerns, participants accentuated the crucial importance of anonymity, desiring a safe space that would not inadvertently expose them to harm.Their anxieties extended beyond mere confdentiality; participants were apprehensive about possible misuse and the potential impact on their familial and social dynamics.These concerns emphasized the multidimensional nature of privacy of domestic workers' records, intertwining economic, cultural, religious, and gender-based barriers, as acknowledged by F1 and F3 respectively in Workshop1: "They are mostly concerned about maintaining the confdentiality of their personal information.Also, whether it will be problematic in their family and social life is another concern of theirs." (F1, Workshop1) "Also, whether it may be misused.So although it may be economically proftable for them in some cases, they are still concerned about cultural, religious, gender-based barriers." (F3, Workshop1) However, this emphasis on anonymity and privacy collided with the necessity of awareness of participants in a virtual meeting, particularly when contextual cues could inadvertently expose sensitive information.The complexity of this predicament was underscored by the example of discussing a family member in a group meeting, unaware that the mentioned individual (or someone close to that individual) was present, highlighting the importance of knowing who is engaged in the virtual group meeting, as F1 exemplifed in Workshop3: "Say in group meeting, one is talking about his brother or wife of his brother, but her brother's wife is present as well in the meeting.It's crucial to know who is in that meeting then." (F1, Workshop3) Additionally, the notion of reviewing recordings before sharing raised a fundamental point -the irreversibility of the shared records.Participants recognized the signifcance of ensuring the accuracy of shared records, understanding that rectifying inaccuracies postsharing would not fully rectify the situation entirely, as F9 shared and domestic workers concurred: "So whenever something is posted or published in social media, it cannot be reversed, as millions of people may have downloaded it already.Thus, it is very important to verify and validate the content by an authorized person before posting anything in social media.Specially, the summary recordings must be validated by the authorized persons." (F9, Workshop3) This notion delineated the distinction between consent for record collection and consent for record sharing, showcasing participants' critical awareness of the potential ramifcations of information dissemination.
Furthermore, participants emphasized the pragmatic value of record summaries over extensive records for sharing purposes.The recognition that lengthy audio contents (both recordings and transcripts) could potentially deter engagement underscored their strategic inclination towards concise summaries, as F8 identifed: "Anyone may not have time to listen to the whole recording.So, the main points may remain unheard because of lost interest beyond the introductory part of the recording." By condensing information into digestible summaries, participants believed they could resonate with a broader audience, thereby fostering enhanced participation and knowledge sharing, as F1 acknowledged the importance of high-quality summaries: "Summary is important as it saves a lot of time and physical energy.However, its usefulness will depend on the presentation of the summary -how well the summary is created." (F1, Workshop3) 4.3 Record Amendments 4.3.1 Participant Agency.The discussions around record access and sharing in Workshop2 and Workshop3 prompted inherent follow-up discussions around record amendments considering various factors.The discussions unveiled key challenges regarding domestic worker participants' agency and involvement in the record amendment processes.Notably, the foremost critical challenge in this regard pertained to the domestic workers' level of awareness of their recording status during their engagements with NGOs and its consequential impact on their agency.The lack of prior awareness of domestic workers about being recorded created a hurdle when participants wished to revisit and review past interactions, ofering limited agency.DW2 revealed this fact in the second activity of Workshop3: "Like we know today that we are being recorded.So, when we come next (maybe after months), we may want to listen to the recordings of this date.But that's because we know that recording is happening today.However, earlier, we never knew whether we were recorded or not.So, there is no question of asking of the recordings." (DW2, Workshop3) This limited agency extended to participants' control over amending inaccuracies in the recorded content.The existing practice positions them as passive recipients rather than active contributors who can efectuate amendments.Consequently, participants must rely on the intervention of responsible personnel to address discrepancies, highlighting the need for an inclusive amendment process, as DW4 expressed in Workshop2: "We'll go to the responsible persons who did this recording.We'll request them to correct it, as we cannot do that." Amid these challenges, participants expressed an aspiration to engage through trusted representatives to address practical issues including record amendments.Collaborating with representatives, whether peers or staf members, was seen as a strategic approach to overcome barriers such as literacy constraints and ensure the accurate records.This collaborative strategy resonates with participants' aim to ensure efective utilization of records, as refected in the statement of DW4: "We can keep this to a trusted staf or representative, so that we can say that we have worked on this project with those staf even long after the project ends (like referencing for later communication)." (DW4, Workshop3) 4.3.2Amendment Process.Participants underscored the importance of establishing record amendment mechanisms within recordkeeping practices, a realization that stemmed from the acknowledgment of the inherent potential human errors.This understanding resonated strongly with the consensus among participants, as F6 noted, "Yes, we should allow this correction.Humans can make mistakes".This realization prompted the participants to provide concrete examples where errors in information dissemination necessitated timely amendments, particularly for addressing potentially harmful misinformation or incorrect observatory note-taking.For instance, DW8 expressed concerns regarding incorrect note-taking during sensitive health-related conversation in Activity1 (familiarization with recordings) of Workshop2, while F9 elaborated on misinformation in Workshop3 by stating: "Yes she should be able to access these notes because if they wrote any wrong information like gastric pain instead of chest pain, or [cracking a joke] say, wrote my [made-up] name incorrectly as Jhoomur [an ornament for head] instead of Nupur [an ornament for legs]; that's not accepted."(DW8, Workshop2) "Say, you got fve training sessions, but you say in the recording you got 6 training sessions, then isn't that a mistake?Questions will arise where you got the budget for the sixth training.Or, you say you attended fve sessions, where you attended four." (F9, Workshop3) The participants also emphasized the need for versatile amendment mechanisms in Workshop3.This perspective was rooted in the conviction that individuals directly engaged and informed about the content were best positioned to report inaccuracies and initiate the amendment process, as echoed by DW4: "The person whose recording contains incorrect information should report the relevant authority by themselves to fx the incorrect information in the recording." Furthermore, participants stressed the proactive nature of such measures in maintaining accuracy, as highlighted by F6: "Regarding correcting the recording, if I (or the person dealing with the recordings) hear this incorrect information (like checkup required after every 6 months for a pregnant woman instead of monthly checkups), I will re-record that with correct information.".
In terms of the practical implementation of amendment mechanisms, participants demonstrated a strong awareness of technological possibilities.They proposed practical suggestions for accomplishing amendments leveraging digital technology including telephony-based solutions, with F2 proposing: "Can it be like this: the person who gave the wrong information can call a number to inform this or rerecord the correct information.Then, the moderator can access that re-recording to make proper amendment of that recording before sharing next." (F2, Workshop3) Drawing insights from existing practices, F3 proposed the integration of coded interactions to streamline the amendment process: "Like we use code number in reports, when participants dial in the correction over a call, then the moderator can fx the original recording using something like a code number."However, the translation of conceptualized amendment mechanisms into reality encountered challenges rooted in existing organizational structures (e.g., limited resources and hierarchical constraints).Participants voiced concerns about the feasibility and potential costs of implementing such mechanisms, underscoring the intricate nature of operationalizing amendments.In this regard, DW7 raised a pertinent query: "Even though we should [have the ability to request amendments be made], will that be feasible, as it may be very expensive!",while F7 highlighted practical operational barriers within NGOs: "My administrative environment will require informing my line manager frst, who will have a fnal decision on how to fx this." A crucial aspect for efectively implementing amendment mechanisms entailed active engagement of stakeholders and the process of consensus-building.Participants fervently advocated for a collaborative decision-making process that involved all relevant parties in the decision-making, as DW2 mentioned: "If someone requests for any change that should be decided discussing with everyone who attended the meeting, as everyone is involved now." (DW2, Workshop3) This sentiment resonates with the overarching perspective that decisions regarding amendments should be collectively made, considering the perspectives of all stakeholders present.4.3.3Communicating Amendments.The discourse also shifted its focus towards the theme: ensuring transparency in the process of incorporating amendments into records.For instance, participants agreed that individuals consuming the audio recordings should be informed about any modifcations that have been introduced.DW3 notably endorsed this perspective, emphasizing the necessity of keeping audiences well-informed.Expanding on the idea, the discussions included the appropriate manner in which the nature of amendments could be communicated to audiences.Participants proposed the inclusion of an introductory verbal description of the changes made at the outset of the recording, as F5 suggested: "It can play a message at the beginning that this has been changed for whatever reason.".This suggestion resonated with the participants, as it aligned with their collective vision of enhancing clarity and transparency for general audiences of their recordings.
Another signifcant theme in the discussions was the perceived value of retaining access to the original versions of recordings.In this context, while DW1 indicated that the edited version sufced for their preferences ("I think only the edited message is fne."),F4 argued that the original message should be entirely deleted due to potential harm ("I think the original one is better not being heard again as it could be harmful, should be just deleted.").
The participants also collectively acknowledged the complexities that might arise from managing diferent versions of recordings, and ultimately reached a consensus that maintaining the original versions held limited value compared to the potential complications it could introduce -a potential need for technological support.This consensus underlined the pragmatic approach preferred by the participants in addressing records' version management issues, as F4 noted: "But it may be problematic for one person only.Okay I get it, since the original message can be harmful or misleading for some, it is better to play the edited one only.Getting a bit complex here." (F4, Workshop3)

DISCUSSION
We discuss key considerations, informed by our fndings, for developing participatory recordkeeping processes in NGO facilitated community voice-based engagements and present design implications for voice-based platforms that facilitate these processes

Designing to Support Stakeholders' Agency
Supporting stakeholders' agency in recordkeeping of community engagement records is crucial to enhance the efcacy and fairness of the recordkeeping processes [22,85].Our fndings reveal that current practices often exclude community members from the recordkeeping processes, rendering them unaware of the existence of records pertaining to their engagement in the process.In many cases, community participants remained uninformed about the collection of records, let alone being actively engaged in their management.This lack of transparency and participation in recordkeeping not only undermines the trust between organizations and stakeholders [7,23] but also stimulates skepticism about weaponizing participation of the marginalized, hampering the potential benefts of efective recordkeeping [85,86].
We fnd stakeholder participants in community engagement activities themselves constantly recognized the inherent value in having access to their own records, and expected a plausible level of involvement in decisions around these records including record access and sharing.This agency over their records is a means of empowerment for the stakeholders, as these records represent their contributions, experiences, and voices -linking back to the original questions around recordkeeping: "why, how, and under what circumstances" [26].We fnd community participants aspired to use these records as evidence of their activities and to deter potential misuse of their personal information, addressing their socio-cultural barriers including household relationship dynamics (e.g., "cannot be shared with our sons", "raise questions to my husband", "husband is at home").Excluding them from the recordkeeping processes can inadvertently drive them to devise alternative approaches, potentially leading to unintended consequences (e.g., DW1 opting for self-recording covertly).Facilitators, on the other hand, needed these records as proof of their work or as references for institutional purposes like securing funding for future projects.However, if facilitators and their donors adopt the practice of documenting all interactions with marginalized communities, it may lead to problematic participation including participation bias [102], which may raise critical concerns like weaponizing participation.
To address these challenges and support stakeholders' agency efectively, we argue that it is imperative to develop recordkeeping processes that demonstrate transparency to stakeholders by keeping participants explicitly aware of, and actively involving them in each stage of the recordkeeping processes from collection to disposal of records [34].This can be achieved through either directly engaging all stakeholders in decision-making, or representation of stakeholders via their trusted delegates.The latter approach helps overcome organizational and literacy barriers that might otherwise hinder meaningful stakeholder engagements [30].We argue that such a shift towards more participatory and inclusive recordkeeping practices can be potentially facilitated by leveraging voice-based platforms (e.g., IVR technology [67]).For instance, selection of trusted delegates can be done by navigating an IVR-menu by button presses by stakeholder participants: '1' for voting for a delegate or '2' for nominating yourself; [if pressed '1'] '1' to vote for 'X' or '2' for 'Y'.At record collection stage or during a live session, participants or trusted delegates can signal to stop recording by pressing a button over the phone.These button press events should be captured by the system, serving as a reference for the participants to tackle weaponizing their participation.A similar approach can be employed regarding other decisions around recordkeeping, including sharing and retention of records, by customizing IVR-menu accordingly or synchronous interactions (e.g., polling) as discussed next.

Designing to Support Difering Perceptions
Our fndings highlight diverging views and motivations of stakeholder parties, particularly in terms of ethical recordkeeping, reliability and security of storage devices, record amendments, and record retention.Stakeholder community participants emphasized the need for ethical considerations, like clear consent, transparency, and confdentiality regarding recordkeeping.Facilitators prioritized record collection mainly motivated by organizational interests (e.g., F2 highlighted monitoring purpose), sometimes at the expense of community participants' comfort (e.g., DW2 shared concerns like being uninformed, while DW5 highlighted organizational fnancial motivations) and negatively infuencing engagement.We suggest that voice-based platforms can streamline informed consent processes utilizing IVR technology [65,67], aligning community ethics with organizational imperatives.
Both parties exhibited reservations about digital storage, citing device reliability and theft concerns.Voice-based platforms, necessitating active role of NGO facilitators, can reassure participants through standard cloud-based security measures [50,80], clarifcation of security measures using IVR (e.g., consent message) [75], and auto-transcription (speech-to-text [104]); thus, enhancing trust relationship between both parties [25,45].However, we argue that this perception regarding vulnerability of digital records to theft of devices i.e., preference for hard copies, indirectly implies that participants deemed these records less valuable than their storage devices: analogous to possessing a valuable jewel but leaving it in an easily accessible glass case [98].This perception underscores a notable gap in comprehending the potential value of these records to distributed communities (i.e., third parties) due to literacy concerns [91], thereby (re)emphasizing the need for proper consent mechanisms that can clearly communicate participants about the broader signifcance of their engagement records beyond their personal use.The consent mechanisms should consider distinct consents for record collection and record sharing.For instance, voice-based platforms can facilitate follow-up asynchronous engagements for taking consent from participants (directly or through delegates) about widening record access to third parties (e.g., creating IVRmenu [65,67]: press '1' to share with family members only, '2' for public sharing, and '3' for yourself only).
Additionally, community participants desired for participatory mechanisms for record amendments and record retention.NGO participants acknowledged the value of amendments and longer retention; however, raised practical concerns such as, requiring approval by higher authorities and associated time and costs.We propose that voice-based platforms can ofer a time-efcient and costefective solution by streamlining this process, where participants can frst record their amendments and send them through using a pre-communicated 4-digit pin, as preferred by stakeholder participants in contrast to existing asynchronous platforms [65,67].Subsequently, amendments can undergo an approval process through either asynchronous [67] or synchronous [95] IVR.Finally, semiautomated editing with artifcial intelligence can ensure their validity and usefulness.For instance, synchronous IVR can facilitate live polling with participants, where participants in a live session can frst listen to the question i.e., consent message/proposed amendment either communicated verbally by facilitators themselves, or played as an audio recording.Then, they can respond by a button press e.g., '1' for approving or '2' for rejecting; [if related to amendment] followed by an automatic or customized notifcation voice-message automatically appended with the approved amended recording to maintain transparency and clarity [96].Addressing usable security [93], the 4-digit pin (potentially, auto-generated) should be distinct to each engagement session; serving as a password to request amendment for a particular session.Additionally, to avoid any call cost for stakeholders particularly, for marginalized communities, the platforms should employ a callback mechanism [48,95].
Furthermore, communities perceived record deletion as a missed income opportunity (e.g., domestic workers' urge for longer retention to avail future projects).In contrast, NGOs valued records as a form of institutional memory, aligning with an analogy between memory and recordkeeping [61].They adhered to rigid institutional deletion rules (e.g., "should be", "ofcially" project's duration or 6/7 years): echoing the concerns related to conventional recordkeeping [6].Voice-based platforms can provide interested stakeholder participants with options to export their records either as recordings or printed transcripts before deletion: a decision to be reached upon synchronous discussion among all participants (e.g., structured group call [48,108]).This decision can also be made by button presses over synchronous or asynchronous IVR similar to consent and amendment processes exemplifed above.These approaches can bridge the gap between organizational protocols and community expectations.

Facilitators' Role to Support Trusted Environment
While voice-based platforms can be a prominent tool in participatory recordkeeping, their efectiveness is realized when complemented by the facilitators' supportive role (e.g., removing sensitive information from recordings [39], facilitating synchronous interactions in a group call [95,107]).We argue and discuss how facilitators' role is crucial, particularly for voice-based engagements and recordkeeping processes.For instance, maintaining anonymity in voice-based discussions (where everyone is invisible) can be challenging, potentially impeding sensitive conversations, and complicating the recordkeeping processes [40].Participants lacking context (e.g., who is speaking) may struggle to contribute meaningfully [16], while revealing context during recording may raise privacy concerns.Facilitators need to apply human judgment to address this intricacy by carefully introducing participants before discussions begin with recording paused, considering the context of the discussions.By actively managing participant engagement, supported by synchronous platforms' functionalities like hand raise and mute/unmute by button presses [95,108], facilitators can enhance the efcacy of voice-based platforms in participatory recordkeeping.Specifcally, facilitators are key to efective consent mechanisms including: tailoring consent messages to the cultural and linguistic nuances of marginalized communities [66]; taking separate consents before record collection, sharing, amendments, and deletion; and designing appropriate IVR-menu.Moreover, facilitators must actively advocate and inform design recommendations for voicebased platforms to align with participants' expectation towards collective decision-making regarding their record amendments, sharing, and deletion.We also argue that ethical concerns, such as self-recording due to participants' perceived insecurity and skepticism in existing recordkeeping practices, can be mitigated through clear consent mechanisms (specifying security measures and access), fostering a trusted environment.Furthermore, our fndings underscore the paramount importance of privacy and confdentiality of records, particularly when broaching sensitive topics (e.g., family-related afairs).Facilitators can enhance privacy and confdentiality by temporarily pausing recording to create a secure space for such discussions.Besides, lengthy audio recordings were deemed overwhelming by participants, making them difcult to review.Additionally, some participants preferred supplementary record formats like physical transcripts for reliability, accessibility and ease of reference.These realization and preferences underscore facilitators' critical role in post-engagement content creation (e.g., summaries, podcasts) and moderation, potentially leveraging voice-based platforms integrated with artifcial intelligence technology (e.g., audio summarization, noise and silence removal [36,109]).
Overall, facilitators mediate between participants and voicebased platforms, ensuring dynamic consent, context, alternative mediums, content creation, transparency, and fexibility in participatory recordkeeping.Facilitators' efciency will ensure the efciency and inclusivity of participatory recordkeeping, ultimately leading to more engaging, productive, and sustainable outcome-based community engagements [43,108].
We acknowledge that there are inherent complexities involved when working in environments with power disparities [46,54,77].As such, our paper does not mandate a rigid framework for participatory recordkeeping approaches including the consent mechanism.Instead, it ofers ethical considerations and strategies for voluntary adoption by organizations.For instance, we exemplify the use of IVR technology, anticipating to signifcantly curtail the challenges of manual consent process in power disparity contexts, as argued by previous research as well [48,81].By encouraging organizations to incrementally implement these practices, our fndings strive to foster a culture of transparency and respectful community engagement over time, and specifcally enhance awareness among marginalized communities about their records.

CONCLUSION
In this paper, we discuss our study carrying out a series of stakeholder focus group discussions around recordkeeping practices in community engagements with domestic workers and NGOs working to support them.We contribute valuable insights from these stakeholder engagements, proposing development of communityoriented participatory recordkeeping processes, particularly tailored for utilization with community-NGO voice-based engagement platforms.We especially emphasize the importance of supporting stakeholders' agency through transparency and active involvement.In addition, we address the need to design for stakeholders' divergent perceptions by advocating for streamlined consent, security assurance, and collective decision-making.Furthermore, we highlight the crucial role of facilitators in maintaining a trusted environment for participatory recordkeeping utilizing voice-based platforms, by ensuring efective consent mechanisms, contextual understanding, fexible access to records, and privacy and confdentiality.We also suggest facilitators can play an important role in content creation and moderation, as well as being able to support enhanced community engagement and knowledge acquisition.This is possible through better utilization of community engagement records.We anticipate that our contributions will improve the efectiveness of participatory recordkeeping in voice-based engagements involving communities and NGOs, resulting in increased collaboration, productivity, and (ultimately) community empowerment.

Figure 1 :
Figure 1: Stickers representing common record types: Each record type was written both in Bengali and English along with a visual cue

Figure 2 :
Figure 2: Workshop Activities: (a) Labeling post-it notes with record type stickers by facilitators during Activity 2 in Work-shop1.(b) One of the group discussions among domestic workers and facilitators during Activity 1 in Workshop3 [permission to use images was granted by participants]