How much should one forget? How much should one remember? These are not trivial questions, if only one could decide. Memory is a key component in systems prone to regime shifts, systems where little things can make a big difference.
Landscapes have the footprint of the past that determines in part the possible pathways the the evolution of the place can take. In ecosystems the frequency of events like fires, floods or outbreaks determine the possible future states of a system, that in turn feed back the frequency of such events. Let’s take as an example forest. An old grown (primary) forest has a low likelihood of fire, however when it happens it’s big due to the long accumulation of fuel. Grasslands, on the other hand present more frequent fires than in turn favor the growing of grass species. Intermediate frequency of fire can trap the system into a secondary forest. Memory here is captured by the type of biomass accumulated, which in turns remember when is more appropriate for fire to come back (and its magnitude) in order to reinforce such biomass configuration.
But when you read the word memory, the first allusion you have is about your own memory, right? Did you forgot the name of this teacher at school or doing laundry last weekend? Retrospective memory helps you to bring information experienced in the past while prospective memory helps you to remember the future, events that hasn’t happen but you don’t want to forget, like your next appointment with the doctor. Humans also have memory and it isn’t only stored in the brain, memory is also stored in their genes, in their social and cultural heritage. For example, a fisherman might make his decision on whether cooperate or not partly based on recent cooperation levels of his fellows, or at least his perception. In the same fishing community, groups with kinship ties that have traditionally worked together are more likely to continue doing so. In such setting, you better follow the rules because bad manners wont be forgot easily.
Now imagine you are a modeler trying to capture such behaviour in your computer. How would you decide how much your agent should remember? Even if you design different types of agents, some whose incredible memory allows them to remember the whole game, and others like me whose memory sucks and need blogs to remember things. How would you decide then on the right distribution of good-bad memory agent types? If you have read till here, you probably already noticed that the memory of the agents will be influenced by the past of the system as a whole (or its perception), which in turn will update their strategy to change the system in the future. Thus, there are nested memories. If you interchange fishermen for lumberjacks, the aggregate behaviour of the jacks will change the memory of fire, which in turn will change the composition of the forest, and likely, their willingness to cooperate given the scarcity of wood. Their strategies would change or not depending their individual memory and the influence of the memory of the system.
Too much memory can trap the system in a rigidity trap. Shakespeare nicely captured it in Romeo and Juliet. Capuleto and Montange families hate was so deep that they couldn’t forgive each other, they couldn’t forget. As a result continuous cycles of disputes made the love of young couple impossible, leaving as last alternative death instead of happy ending. Too little memory, on the other hand, left the system without a pool of already existing components, the Lego bricks for new strategies to develop. One cannot create a new recipe without knowing what kind of ingredients are available in the area, which ones goes well with each other, what are the cooking times and which wines could match better the expected flavor. In other words, you cannot start from scratch, one usually start from fragments of other recipes to avoid your partner throwing you the plate on the face. Lack of memory renders the system without creativity. So better forgive your enemies and don’t forget the basics of cooking.
These ideas were inspired by a number of people I’d like to acknowledge. Steve Lansing for inspiring discussions on the irrigation systems in Bali, where apparently kinship is shaping cooperation at the community, and memory or lack of it plays a role at identifying regime shifts nested in scales. Jamila Haider for sharing her ideas about food traditional systems in central Asia. She is looking at how food traditions have been shaped by the environment both political and ecological, and how current changes in agricultural practices in turns are shaping both the landscape and the local gastronomy. Scott Heckbert with who I took a course recently on agent based modeling and helped me to think on irrigation models. Jacob von Heland who recently defended his PhD on morals, culture and resilience. I got captured by this passage:
Rowing backwards into the future is the full title of this thesis and it has become suiting over time. Rowing looks simple but requires some skill, as well as figuring out relations to help determine which direction to go and who and what to bring along. When at sea it is necessary to make do with what there is and while the boat is always moving the oars, stars, senses, birds, smells, tools and sharing work with company give a good chance of getting your motley crew somewhere you want.
Jacob is part of this group of people who likes to look at the past to better understand the future. And last but not less, Ana Lucia Cardona who is trying to detect early warnings (diagnosis) of mental diseases by looking at slightly changes on the prospective memory (the memory of the future) of patients in early stages of dementia. Which brings me back to the initial question: How much should I forget? How much should I remember? If only one could decide…