<aside> ⬆️ back to The ecosystem of 'social agriculture'
</aside>
In this section:
Social agriculture depends upon the logic of social media platforms, which largely depend on ad-driven business models. We can't cover the full extent of algorithms and features that power these social media platforms here, but we mention a few examples that have a clear bearing on social agriculture below:
We start with the most basic example of "temporal messaging queues". This underpins some of the major platforms used for social agriculture, e.g. WhatsApp and Telegram.
We don't need to get into technical details of how these queues are implemented. All we need to understand is that messages are brought to a viewer's attention based on the time they were sent.
A diagram of how basic messaging queues work
People often seek this predictable design in social agriculture — the temporal messaging queue reflects an intuitive model of everyday communication, i.e. the message in pole position to grab a user's attention is simply the last thing sent. These models are intuitive, predictable and easy for users to understand as everyday communication utilities. Tools that use this messaging paradigm become critical for everyday agricultural practices, e.g. sending instructions to a farm manager, or taking orders from customers. We highlight this model mainly to contrast it with the very different model of 'algorithmic news feeds', which we look at next.
Most social media platforms employ some form of News Feed feature. This acts as a personalised front page for every user, where information presented in pole position to each person is carefully orchestrated by algorithms. Platforms like Facebook and Douyin use this model.
.webp){kind=logo}
News feeds are critical to these platforms' advertising business models. The way these feeds work is opaque, as compared to the simple messaging queue model outlined above. These systems required complex software architectures from their earliest incarnations (Facebook first released the news feed in 2006). The video below explains how Facebook's current news feed is powered by machine learning to score posts based on "signals" (such as user likes). The video claims this work enables Facebook to personalise the news feed for every user with "the content that matters most to them".
.gif)
[source: Facebook](https://www.facebook.com/plugins/video.php?href=https%3A%2F%2Fwww.facebook.com%2FEngineering%2Fvideos%2F264352435037706%2F&show_text=0&width=560)
source: Facebook
There are reasons to doubt that Facebook's algorithm straightforwardly personalises our news feed to show us posts that matter most to us. The companies business objective is better described as maximising "time on device", a term taken from the gambling industry to measure the performance of slot machines. In fact, news feed algorithm design has been inspired by concepts from slot machine design, such as irregular, or intermittent reward schedules to keep users hooked. These algorithms maximise attention not content value from a user perspective.
Intermittent reward mechanisms generate more "time spent playing", play with a great interactive demo to learn more
Time on device is misaligned with the requirements of better information exchange — If such news feed algorithms are designed to maximise time on device, this is problematic for social agriculture. In our analysis Examining large scale groups in social agriculture in Kenya we found that information sharing in large agricultural groups accounted for a significant share of posts. Moreover, when we asked agricultural workers about The experience of social agriculture from users in Kenya, they reported that "getting farming tips" was their main activity, giving us reason to believe that information exchange observed in groups was "what matters most" to users.