PaperlayerView exact revision on arXiv

Source-linked AI summary

Botivist: Calling Volunteers to Action Using Online Bots

Saiph Savage, Andres Monroy-Hernandez, Tobias Hollerer

arXiv:1509.06026v1cs.SIcs.CY

TL;DR

Activist groups need casual volunteers, yet social media offers little direction and campaign messages can be difficult to test. Botivist uses Twitter bots to probe different calls to action; its direct strategy yielded a reply rate over 80% and 94% relevant responses.

Abstract

from arXiv · show

To help activists call new volunteers to action, we present Botivist: a platform that uses Twitter bots to find potential volunteers and request contributions. By leveraging different Twitter accounts, Botivist employs different strategies to encourage participation. We explore how people respond to bots calling them to action using a test case about corruption in Latin America. Our results show that the majority of volunteers (>80%) who responded to Botivist's calls to action contributed relevant proposals to address the assigned social problem. Different strategies produced differences in the quantity and relevance of contributions. Some strategies that work well offline and face-to-face appeared to hinder people's participation when used by an online bot. We analyze user behavior in response to being approached by bots with an activist purpose. We also provide strong evidence for the value of this type of civic media, and derive design implications.

INTRODUCTION

Botivist addresses activists’ difficulty recruiting casual volunteers and testing campaign messages by using online bots to try different calls to action. Its Twitter experiment found that message strategy affected participation, with direct calls prompting the most contributions.

INTRODUCTION

  • Activist groups need casual volunteers, but social media offers little direction and groups struggle to predict or test which campaign messages will mobilize people.Groups may rely on a social-media point person, yet transferring their knowledge is difficult and testing campaign variants can be costly or impossible.Source: INTRODUCTION · pp. 1–2
  • Botivist lets activist groups try different strategies for requesting contributions from strangers, reducing time spent on people who may never contribute.Activists provide a cause, and the platform tests message strategies on social media.Source: INTRODUCTION · p. 2
  • The paper examines how users respond to differently framed activist-bot messages and studies the communication patterns of those who engage.The authors note that engaged users could potentially become core activists.Source: INTRODUCTION · p. 2
  • The experiment used Spanish-language Twitter bots to invite collaboration on corruption and impunity, with calls structured around three mentioned users.The researchers chose Spanish and the Latin American corruption case based on their language and activism experience and access to Mexican activists.Source: INTRODUCTION · p. 2
  • 175 volunteers responded and made 1,236 contributions; direct calls prompted the most contributions, while less openly persuasive messaging encouraged almost double the replies.The contributions comprised 424 tweets and 813 favorites and retweets.Source: INTRODUCTION · p. 2

Botivist: Strategically Framed Messages

Botivist applies message-framing research to activism by testing alternative calls to action with online bots. It is designed to probe strategies rather than assume which framing will work.

Botivist: Strategically Framed Messages

  • Botivist gives organizers flexibility to audience-test different calls to action, addressing a gap in persuasive-computing research on activism.Prior systems often coordinated volunteers without testing alternative message framings.Source: Botivist: Strategically Framed Messages · p. 3

Botivist: Online Automated Agents

  • Botivist builds on work showing that simple online agents can acquire social relevance, using bots to mobilize people for activist groups.Earlier bot research focused more on detection than on persuasion or mobilization.Source: Botivist: Online Automated Agents · p. 3

BOTIVIST

  • Botivist probes different ways of presenting a call to action because the framing needed for a desired turnout is not easy to decide in advance.The system is a web application for activist groups addressing social issues.Source: BOTIVIST · p. 3

Botivist Strategies

  • Botivist uses separate Twitter accounts for its strategies; the accounts identify as bots and otherwise share the same main characteristics.Messages were based on common activist strategies and polished with activists’ help.Source: Botivist Strategies · p. 3

Direct Strategy

Botivist compares direct, solidarity, and gain framings for calls to action. The strategies differ in how they address participation, from a straightforward request to adding an appeal or benefit.

Direct Strategy

  • The direct strategy explicitly asks people to act together to find solutions to a social issue.Its sample call invites people to brainstorm solutions to corruption.Source: Direct Strategy · p. 3

Solidarity Strategy

  • The solidarity strategy pairs the direct call with a separate solidarity quote intended to invoke solidarity or empathy.It sends two tweets for each call or reply: the request and the quote.Source: Solidarity Strategy · p. 3

Gain Strategy

  • The gain strategy adds a benefit people might receive by participating to a call otherwise modeled on the direct strategy.The example says participation might improve cities.Source: Gain Strategy · p. 3

Loss Strategy

Botivist’s loss strategy frames nonparticipation as a potential loss, while its workflow identifies potential volunteers, invites them to act, and follows up on replies.

Loss Strategy

  • The loss strategy frames a call to action around what people might lose if they do not participate.Its example warns that cities might suffer without collaboration on solutions to corruption.Source: Loss Strategy · p. 3
  • Botivist is intended for early collective-project work, prompting discussions and plan proposals through “How Might We” questions.Replies are randomized, each person is targeted only once, and the system engages further only after a response.Source: Loss Strategy · p. 3

Botivst Operation

  • Activists define the social problem, after which Botivist identifies and tweets potential volunteers; it targets equal numbers for each strategy.Figure 2 presents the workflow for calling volunteers and obtaining responses.Source: Botivst Operation · pp. 3–4, Figure 2 · p. 4
  • Botivist finds people who recently tweeted relevant keywords, randomly assigns them to strategies in groups of three, and monitors for replies.Managers can supply keywords or hashtags; follow-up messages depend on the assigned strategy and continue prompting action.Source: Botivst Operation · p. 4

EVALUATION

The evaluation asks whether online bots can recruit volunteers for social problems and elicit relevant proposals or discussion. Botivist was tested publicly on Twitter with calls about corruption and impunity.

EVALUATION

  • Table 1 presents the questions used in the Direct strategy.Source: Table 1 · p. 4
  • The evaluation tests whether bots can attract volunteers and prompt relevant contributions, including discussion or proposals addressing a social problem.It also asks which solicitation strategies are most conducive to obtaining responses.Source: EVALUATION · p. 4
  • Botivist publicly called volunteers to action about corruption and impunity on Twitter.The study reports attracting 175 volunteers over two days.Source: EVALUATION · p. 4

Method: Botivist and Volunteers’ Participation

Botivist alternated calls about corruption and impunity across four strategies and tracked volunteers’ replies and interactions. Direct calls mobilized the most unique contributors, while other measures of engagement varied by strategy.

Method: Botivist and Volunteers’ Participation

  • From April 26 to May 7, 2015, Botivist alternated calls between corruption and impunity and randomized call delays to reduce the chance of being labeled spam.Researchers monitored the accounts and collected sent and received tweets for analysis.Source: Method: Botivist and Volunteers’… · p. 4
  • 175 volunteers made 1,236 responses or other actions, including 424 tweets, 320 interactions with Botivist content, and 493 with other volunteers’ content.Source: Method: Botivist and Volunteers’… · p. 5
  • Over 30% of people targeted by Direct responded, and this strategy gathered over 50% of participants and mobilized the most unique contributors.Figure 3 summarizes volunteer counts and responses by strategy.Source: Method: Botivist and Volunteers’… · p. 5, Figure 3 · p. 5
  • Direct had an 81% reply rate; gain and loss rates resembled product reviews, while Solidarity’s was 21%.Direct’s rate was the highest across strategies.Source: Method: Botivist and Volunteers’… · p. 5
  • Solidarity received the most retweets and favorites for its content but the fewest volunteer replies, despite significant interaction-rate differences across strategies.Direct and Loss generated the most interactions with other volunteers’ content.Source: Method: Botivist and Volunteers’… · p. 5
  • Direct and Loss generated the most retweets and favorites among volunteers, who interacted more with one another’s content than with Botivist’s.Source: Method: Botivist and Volunteers’… · p. 5

Method: Analysis of Volunteer Responses

The response analysis classified volunteers rather than individual tweets to assess whether they contributed relevant discussion or proposals. Three Spanish-speaking coders independently reviewed the volunteer responses, with a third resolving disagreements.

Method: Analysis of Volunteer Responses

Results

Most volunteers contributed relevant discussion or proposals, but response relevance differed by strategy. Direct calls were associated with nearly all volunteers discussing solutions, whereas Loss drew more off-topic responses, many questioning bots’ role in activism.

Results

  • 81% of volunteers responded with discussion or proposals relevant to their assigned social problem.Source: Results · p. 6
  • Only 19% of volunteers contributed solely to off-topic discussions, where “bots” was the most-used word.Source: Results · p. 6
  • The Direct strategy had the most on-topic volunteers, while Loss had the largest share of off-topic replies.Almost all Direct volunteers discussed or brainstormed solutions; over 60% of Loss off-topic replies used the term “bot.”Source: Table 3 · p. 6, Results · p. 6
  • Loss was more effective than Direct in face-to-face laboratory discussions, but the online pattern was opposite; Loss replies often questioned bots’ role in activism.The authors suggest that bots with an obvious agenda may be seen as suspect or unwanted.Source: Results · p. 6

Method: Communication Patterns of Volunteers

The authors compared responders’ and non-responders’ pre-contact tweets, using ranked terms and qualitative coding to characterize each group. Responders more often used activism-related hashtags and mentions, while non-responders’ terms more often concerned politics and news.

Method: Communication Patterns of Volunteers

  • The analysis compared each targeted user’s 200 tweets before Botivist contacted them, separating responders from non-responders.Pre-contact data were used because Botivist’s interaction might influence users’ later conversations.Source: Method: Communication Patterns of… · p. 6
  • Mann-Whitney ρ ranked words by group-specific frequency per user and highlighted terms used more by one group than the other.The authors selected the top 1% as each group’s key terms; most were hashtags or user mentions.Source: Method: Communication Patterns of… · p. 6
  • Three Spanish-speaking colleagues categorized 155 key terms using mutually exclusive categories, with two coders agreeing on 143 terms (Cohen’s kappa = .79).The categories were developed through qualitative coding and are described with examples in Table 4.Source: Method: Communication Patterns of… · p. 6, Table 4 · p. 7
  • Over 50% of responders’ key terms related to activism, while non-responders’ key terms more often concerned news (40%) and politics (43%).Only 6% of non-responders’ key terms were social-cause hashtags.Source: Method: Communication Patterns of… · p. 7
  • Marketing analytics terms appeared in 4% of responders’ distinct words, compared with less than 1% among non-responders.The authors note that responders mainly used activism-related hashtags.Source: Method: Communication Patterns of… · p. 7

DISCUSSION

Botivist’s direct strategy drew high replies and relevant responses, while results varied with message framing. The discussion highlights audience fit, transparent communication, and design choices for sustaining volunteer collaboration and managing automated outreach.

DISCUSSION

  • Botivist’s direct strategy yielded a reply rate of over 80% and 94% relevant responses.The authors describe free relevant contributions as meaningful because maintaining crowd-based systems is costly.Source: DISCUSSION · pp. 7–8
  • Strategies effective face-to-face did not translate uniformly: solidarity produced the fewest replies, while loss produced the fewest relevant replies.The loss strategy also appeared to make some people question bots’ role in social issues.Source: DISCUSSION · pp. 7–8
  • Responders were involved in online activism and marketing, and people responded more to bots transparent about their purpose than those seeming to have a “secret agenda.”The authors emphasize understanding the communities where bots are deployed.Source: DISCUSSION · p. 8
  • Changing a bot’s message altered response rates without interface changes or complex bot behavior, suggesting dialog design can affect volunteer participation.The authors suggest persuasive technology may contribute to preferences for less manipulative tactics.Source: DISCUSSION · p. 8
  • Botivist could be adapted to social issues with formulable messages, and city planners could recruit experts, stakeholders, or minorities to design more inclusive urban spaces.Source: DISCUSSION · p. 8
  • If activist bots become popular, they may compete for users’ attention; platform APIs could support them, while bots could avoid duplicate requests and recommend more effective strategies.The authors also identify compassion fatigue as a reason to prevent repeated volunteer requests.Source: DISCUSSION · p. 8
  • Botivist’s public mentions helped strangers meet and collaborate, while long-term cooperation may need follow-up, recognition, community-building, or standardized contributions.Automated agents may need to maintain contact without pinging volunteers so often that they tune them out.Source: DISCUSSION · p. 8

CONCLUSIONS AND FUTURE WORK

Botivist’s direct strategy was most effective for relevant participation, and the experiments demonstrate the feasibility of using online bots to call volunteers to action. The findings are bounded by a Spanish-speaking population and a breadth-focused methodology.

CONCLUSIONS AND FUTURE WORK

  • A direct strategy was most effective for relevant participation, while tactics effective face-to-face were less effective when used by online bots.Volunteers also began interacting with one another, showing potential for bots to scaffold collective efforts.Source: CONCLUSIONS AND FUTURE WORK · p. 8
  • The findings may not generalize beyond Spanish speakers, whose cultural attitudes toward bots may shape their responses.The paper notes recent accusations that Latin American governments used bots to influence political discussions.Source: CONCLUSIONS AND FUTURE WORK · p. 8
  • The methods prioritized breadth over depth; future work could use longitudinal studies and in-depth volunteer interviews.Source: CONCLUSIONS AND FUTURE WORK · p. 9

Related papers

Loading 1509.06026v1…