Kitchens participate in the Assembly.
Kitchens are sites where food is processed and consumed, characterised by food rituals and cultures, food preferences, allergies, nutritional and dietary requirements, varying abilities to process ingredients, cook or conserve food. Kitchens feed these conditions into the system by manually entering data manually or deploying sensors. Kitchens participate in the Assembly.
SMUC Research Projects
Externally Invited projects
The Assembly is the collective decision-making body of kitchens.
The Assembly is the collective decision-making body of kitchens. It is a site for negotiating the politics of food distribution (such as fighting hunger, minimising food waste, or building more just food systems). It is also a site for negotiating food items’ situated value in specific kitchens under specific conditions and a fair distribution. The Assembly’s understanding of when an ingredient has the highest value in a kitchen determines its sensing and monitoring strategies to generate meaningful data for the food distribution. The Assembly establishes the protocols of food distribution and engages in Machine Teaching to delegate its orchestration to a distribution algorithm. Data stories shared in Assembly meetings evidence the value of ingredients, the joy of cooking, annoyances and emergencies as an outcome of the collective machine teaching. Data stories are the Assembly’s gauge for how well the infrastructure serves the kitchens and so they can adjust the protocols, datafication strategies, and re-teach the machine accordingly.
SMUC Research Projects
Externally Invited projects
Data labelling is a technique for generating training data for machine teaching.
Data labelling is a technique for generating training data for machine teaching. Labelling assigns specific ‘classes’ to data, so that a model trained on this data will classify future (unlabelled) data input. For example, to assign rescued food items to specific kitchens under specific conditions, labellers assign kitchen names (labels) to data showing conditions in individual kitchens (such as fridge and pantry contents, ripeness of stored food, motivation to cook, volume of hungry stomachs, calories burnt, nutritional needs) and food that is available for distribution (a box of cereal). Maintaining the Assembly’s autonomy over its training data through involving Kitchens’ in collective data labelling is key to just and non-extractivist models of algorithcmis governance of the commons.
SMUC Research Projects
Externally Invited projects
Algorithms thrive on uniform repeating patterns.
Algorithms thrive on uniform repeating patterns. Speculations about machine teaching commons must consider safeguarding different voices and non-intersecting interests in the commons. Machine teaching strategies must embrace good enough prediction (of where food is most valuable in a given moment) instead of ultimate optimisation. This goal-setting is a result of shifting priorities for the available smaller datasets, the Assembly’s increased autonomy requiring minimal feasible datafication, holding space for frictions and renegotiated preferences in the Assembly.