In the 1960s, Abraham Maslow infamously forewarned , “it’s tempting, if the only tool you have is a hammer, to treat everything as if it were a nail.” Technology of all kinds falls victim to this bias, where we over rely on a familiar tool, or are blindsided by the novelty and promise of the new.

Community is an innately human endeavour, with relationships at the core. It’s a living system, like a garden, and while we may need to prune and apply weed killer periodically, taking to it with a hammer would quickly murder the flowers.

So it is with applications for Artificial (or more accurately, Augmented) Intelligence. We need to parse the nails from the flowers and apply the right tool for the job. Machines can help build better online communities, if we’re thoughtful about putting them to work.

How AI can add value to online communities

There are four main areas where AI can assist online communities and the work of online community professionals:

Reduce harm
Automate burden
Organise meaning
Deepen insights

Reduce harm

AI toolsets can shield humans from high-risk or toxic content or behaviour, protecting their wellbeing and the sanctuary of community experience.
• Psychologically harmful content moderation
• Workload efficiencies
• Triage

Those who moderate online communities can witness the worst of humanity. Graphic, violent and other high-risk content has lasting impacts on mental and physical health. Perhaps the most compelling application for AI is to reduce harm to humans exposed to this content in the course of online community governance.

AI tools can offset moderation burden and assist with triage of issues and incidents, ideally reducing the time spent in the worst of the trenches. Content that no human should ever have to see can be swept up by a machine, then if absolutely necessary, reviewed under human supervision in a protected setting, at a pace that considers the psychic toll of the process. This in turn minimises exposure for community users, who might otherwise run into this content, risking damaging personal and social outcomes.

AI can also support reputational workflows that reward desired behaviour within communities (as determined by community architects and participants), reinforcing healthy participatory norms.

It’s important to note that using AI to filter and triage graphic and high-risk content on social media platforms is a very different use case. In hosted online communities, the instance of harmful, toxic and high-risk content is generally much lower, as there is functionality for community managers to vet the membership threshold, configure culture, enforce social norms, and empower self-governance. On Big Social, intentional communities are nested within platform cultures geared toward commercial, not civic, logics, and community managers typically have far less control, lacking tools to manage governance contextually for their users.

Automate burden

AI can alleviate the burden of repetitive and banal community house-keeping, restoring staff and user focus on more desirable or valuable activities.
• On-boarding and orientation
• (Re) engagement nudging
• Low-tier moderation

Automating repetitive tasks is the lowest hanging fruit of machine assistance for online community. Community professionals tirelessly work to articulate and evangelise the depths of their practice and its strategic value. When they’re caught ‘in the weeds’, it is almost moot.

Removing spam, moving content in the wrong place, onboarding newcomers, matchmaking needs with haves, routing issues or opportunities to key users, repeating messaging or processes day-in and out, all reduces precious time we could invest proactively and strategically as digital behaviourists.

Automated on-boarding and prompts around behavioural nudging - making sure new members connect with guidelines and get a tour. Showing them how to make the best first post possible, or matching them with a community member they have things in common with (or are totally different from - depending on the purpose of your community).

Smart automation can increase conversion, engagement and retention; help motivate members to act; and make it easier to recognise and reward desired behaviour. It can also help with triaging risk events and issues.

AI that monitors and automates community house-keeping helps community professionals focus on proactive, strategic social interventions that steer purpose and shared value, and lets community participants focus on the highest value aspects of their experience. It stops them turning to a community team member for minutia or busy-work, and creates space for deeper, exploratory interactions.

Organise meaning

AI tools can help organise and make shared sense of the information and activity in a community.
• Retroactive and pre-emptive tagging and classification of content
• Routing/re-routing to the right area or person
• Clean-up and archival
• Architectural reordering

Sense-making is an activity community professionals scaffold and incentivise, to deepen social ties, cement shared symbol systems, and unlock unique value in the community experience. AI tools could help us with this process, making it easier for participants to organise and make shared sense of community activity and knowledge generation.

This could be as simple as applying tags to posts, suggesting new tags if content is trending, or a new naming convention to meet an emerging pattern. If a hot button issue is emerging, community professionals could train or trigger their machine assistance to create a sub-forum to house discussion around that topic. Machine learning might be capable of suggesting a new function, feature or topic for the community, based on predictive insights. Community architecture itself may become more fluid with the support of machine learning and a community manager to feed its smarts.

This extends to personalisation - a common application of AI in digital experiences. The essential caveat with personalisation is that it must align with the core community use case.

Part of what makes community an effective, binding operational model is shared experience. If personalisation re-silos a person and prevents them from experiencing the commons, this erodes the social value of the community. There is intrinsic value in informational and social foraging within a community, in serendipity and discovery, in witnessing and being witnessed. If these are core to your community value proposition, personalisation should be approached warily. But if personalisation can increase engagement and help participants both contribute and extract value, AI enabled personalisation can be a boon.

Can AI, via analysis and predictive insight, help move community members through optimally rewarding journeys and pathways within the community. Communities are about progress. Participants need to make progress or they will not engage over time. Consider how AI can help users and members draw cumulative value from engagement. Can it be applied to help them level up, achieve mastery in a subject matter or social role, serve them content that challenges their comfort zone?

Deepen insights

AI can capture and organise data sets to reveal new or important insights about our community.
• Sentiment analysis
• Map influence and networks
• Understand content and behavioural trends
• Uncover social roles
• Leading indicators for risks and opportunities
• Risk and event modelling

After automation the most promising aspect of AI is its ability to produce richer insights about our communities. From richer dashboards and reporting capability, to the unlocking of ‘hidden’ risks or opportunities, machine learning promises rich excavation of our community strata.

Data visualisation tools already let us map networks and centres of social capital. With AI we can go deeper. If we teach a machine learning system the criterion for a set of social roles we want to identify in our community - curator, evangelist, moderator, connector, peacemaker, archivist - it can work on mapping who fills those roles now, and who might in future based on current indicators. Are there community champions, subject matter experts, or natural empaths in-waiting we need to meet with support and opportunity?

Community managers currently create manual documentation around member personas, lists of topics members are expert in, and so on. AI that can help a community manager assess and capture this information, and organise it systematically will be a critical sense-making tool for their work, and the community at large.

Are unreported benefits being generated within the community? Users getting tasks accomplished, achieving learning outcomes, making lasting connections. Whatever the use case, robust AI should, over time, help us uncover patterns within the community experience that deserve or require strategic attention.

AI can help us capture and benchmark content and behavioural trends. We can harness predictive tools to model actions or events within the community. What might happen if a key contributor departed - or took on a new role in the community? Is there an equivalent of an earthquake detector for your community, which might help you get a jump on a risk or an opportunity before it catches fire?

Community professionals already perform some of this work, labor intensively. AI may help us do it faster and more comprehensively, so we better quantify and express the value of our work, and focus on the stories that value illuminates.

Download the recording of my Vanilla Forums webinar on AI and online communities

Context coaches

In addition to the substantive ethical issues around AI, such as algorithmic bias, there are other risk factors that can create problems for community professionals and their constituents.

Chief among these is what I call context flattening - where nuance is collapsed and all roads creep toward the same use case, whether appropriate or not. Algorithms are formulas for intention. We have a responsibility to ensure that intention aligns with our community purpose, shared values and goals.

Customer support logics applied to therapeutic forum jeopardise a judiciously cultivated community culture. A reputation algorithm that prizes quantity over quality will at best, incentivise noise over signal, and at worst, hand a trophy to a serial troll.

Community professionals must work with technologists to align intentions and logics with those of your community.

If we’re not cautious stewards of context, AI risks further problematising the work of online community building rather than liberating it.

When to introduce AI to your online community

There’s no single litmus test telling you when to add AI, but the following four qualifiers are a good rule of thumb:

When it reduces harm
When it lifts burden
When it deepens insights
When it quantifies value
When it scales value
When, in doing the above, it does not fundamentally diminish essential human worth or experience.

Orientate with these important questions:

What behaviours, activities or content am I reducing or removing by applying AI to this function, and how important are they to purpose and participant experience?

Where can AI create social or experiential efficiencies for participants that enhance the value of this unique community?

Hero the human

Community builders understand relationships, and how to build sustainable social systems around purpose. You might say we’re building the human infrastructure of the internet. This puts us in a unique position to influence, test and enhance AI systems - to make sure they centre the human and they’re applied equitably and contextually.

AI can be a powerful tool to democratise insights - but we’ve seen how the promise of other democratising technologies has played out. Community professionals, as digital behaviourists, can play a leading role in shaping ethical and purposeful AI that heroes the human and helps us gather more closely, rather than splintering and atomising us further.

Artificial Intelligence for online communities