The growing use of agricultural land at the expense of wilderness is having a profoundly negative impact on our ability to mitigate the impacts of client change. The accelerating clearing of land for farming is driven by a growing consumption of agricultural products, primarily as a result of an expanding and more affluent global population that wants more and better food. Technological advances have increased the calorie yield of farmlands but struggle to keep up with the world’s appetite.
This farmland has to come from somewhere, and more often than not it is at the expense of wild areas. By removing natural carbon sinks, such as forests, to expand agricultural land we are not only increasing greenhouse gas emissions (e.g. by burning woodland or introducing methane producing animals), but are also constraining the planet’s ability to regulate CO2 in the atmosphere by removing the plants that sequester it out of the air. Not to mention the general destruction of animal habitats, removal of natural flood defences and the degradation of wilderness into monotonous fields.
Dietary change is one of the single most impactful things we can do as individuals to positively help the natural environment and in turn put the brakes on climate change. This is something that a lot of people want to do but don’t necessarily know how to without drastically changing their lifestyles.
The question is, how can you change what and how a population eats in order to tangibly reduce the need for agricultural land that can be returned to wilderness? And how can you communicate to people that what they are doing is making a difference? Machine learning (ML) has been proven to be able to impact population scale numbers of people through personalisation of news delivery that has gone on to swing elections, but can it be used to change how people eat?
Where did all the fields come from?
Half of the world’s habitable land, that is not barren or covered in ice, is used for agriculture. That is a staggering amount of land surface about the size of the entire Americas, China and South East Asia. Amazingly that has grown by 270% since 1850. Of the 51 million km2 of agricultural land, 77% of that is set aside for meat and dairy production supplying only 18% of the world’s calorie intake and 37% of the world’s protein intake.
Although the overall land requirement per person is gradually decreasing, population growth is compounding to a net growth in agricultural area. Clearly something needs to change otherwise the world will be fields.
Food consumption is the direct driver of land use. If people eat more burgers, more land is set aside for cattle farms. If people eat more nuts, expect to see rows and rows of nut trees springing up in response. David Attenborough made a point that you don’t see herds of carnivorous predators roaming the African savannah, and that is because the number of herbivore animals the land sustains does not support large numbers of meat eaters. However affluence, cultural norms and habit have turned humans into the world’s largest carnivorous herd, putting undue pressure on the land.
It quickly becomes obvious that the human diet needs to change to free up land that can recover as wilderness if we are to mitigate risks accentuated by climate change.
Where does AI fit in?
Using ML and AI to build recommendation engines is not a novel idea. ML is behind what series Netflix suggests to you, what songs Spotify puts in your Discover Weekly and what news and memes you see in your social media feeds. More specifically in the retail and FMCG sectors it is used to personalise online shopping experiences by customising the content of home pages, generating autocomplete search terms and suggesting supplementary items based on what is already in a customer’s basket.
The typical retail recommendation engines do things like map customer behaviours from search and purchase history, customer profiling by demographic segmentation and collaborative filtering where like-minded customer purchases are suggested. In FMCG, particularly food retail, online data can be combined with in-store purchase history linked to individual customers via reward schemes to build up a more complete customer profile. This in turn is used to further refine the recommendation engine and maximise the sales conversion of suggested products. Well delivered personalised in-store and online shopping experiences, combined with customer loyalty schemes that distribute vouchers and rewards, can be very powerful tools to influence how people shop.
So how can recommendation engines be used to change how people eat and as a result reduce agricultural land area?
As mentioned previously, land use has been shown to be directly attributed to consumer behaviour, behaviour that can be shaped and changed by recommending certain products and not others. I propose a recommendation engine based on customer profiling and purchasing history that suggests product substitutes with a lower land area requirement than previously purchased items.
Over time the objective of these recommendations would be to steer consumers towards lower land use alternatives, and generate eating habits that develop into the medium and long term of a customer’s lifecycle. This could be done overtly if, for example, a customer chose to have less land intensive substitutes suggested to them, or more subtly by changing how customers see products both in store and online, or by providing vouchers for more environmentally friendly products.
There are a number of key success factors, in addition to a well crafted recommendation engine, that are required to make such an idea have its desired impact.
Collecting land use SKU data
Being able to suggest lower land use products requires knowledge of the land use requirements of all products in the first place down to a stock keeping unit (SKU) level. This data would be relatively easy to obtain for simple products such as oats, milk and apples but gets seemingly more complicated when products like cakes, with multiple ingredients, are involved. As far as I am aware no dataset like this exists, and will be both time consuming to collate and to verify its accuracy, however it is essential to providing legitimate and impactful recommendations.
Understanding why people buy certain foods
Why do people eat the foods they do? Is it because they like the flavour, texture or nutritional value? Is it a sign of status, or culture, or just sheer habit? Developing a fundamental understanding of why people eat what they eat will be essential in suggesting substitutes that are actually valid. A Beyond Meat burger is going to do a better job than a big mushroom if I’m really into my beef patties. Crowdsourcing feedback on the effectiveness of substitutes as they are recommended will enable incremental improvements to the quality of recommendations over time.
Incentivising people to keep up behaviour
Behaviour has to be sustained over a period of time to make a real impact and for this a mechanism for incentivising customers to keep buying lower land use alternatives would need to be concocted. This could involve providing metrics showing individual positive impact over time, gamifying the process by introducing badges (earn a badge for every 10 fields you re-wild) and providing information on other customer behaviour and impact to create a “we’re in this together” mentality. It’d be cool to know that all the people who shop at your local supermarket contributed to reforesting a national park equivalent area over the course of a year. The same incentives would need to be extended to retailers who may be sacrificing margins by selling more cheaper products (e.g. selling beans instead of steak), however providing accurate metrics on impact would allow for external reporting which may grow a more environmentally conscious customer base.
If implemented at scale I genuinely believe that this type of recommendation system could fundamentally change how populations eat and as a result reduce global agricultural land requirements. Re-wilding that land with forests and grasslands to its former glory could help to mitigate the impacts of climate change. Yes there will be some people that don’t buy into it, but by informing conscientious customers through providing information on alternative products and the tangible impact that their decisions are having, I believe enough people would get on board to make it worth doing.
This for now is merely an idea and a few open questions remain which might be the focus of future articles or work, who knows:
- How can you collect accurate data on the land requirements of all products at an SKU level, and how can you verify that?
- How can you communicate with customers the impact that their decisions are making, and how can you help people to understand and contextualise that impact?
- How can you educate people on the benefits of reducing the land area required for agriculture?
- How can you persuade retailers to implement the solution when the products that are recommended may lead to lower margins?
- How can you influence people who don’t care?