Agentive Apprenticeship

Where does artificial intelligence fit with learning? In terms of the future of learning with Artificial Intelligence, it really comes down to a decision of privacy and who owns your data. As we have discussed open is a set of practices ghat do not have to conflict with privacy. Yet if you do want to leverage AI, controlling the data you create will allow you to have a say on the learning process. We need to think of agentive apprenticeships as a method of returning learning to a state of holistic technologies. We saw in our readings on Artificial Intelligence that Large Language Models have convergence problem when compared to mainly by requiring each person to have their own place online and then by creating networks of community and knowledge through affinity spaces. ### Agentive Apprenticeships and Open Pedagogy Our views of community, knowledge, learning and technology plus our participation on the web has shaped the principles we build towards through open pedagogy. Innovation through Community, Knowledge, Learning and Technology creates the ability to use education as an innovation system. Open pedagogy can act as an innovation system. According to Freeman (187) Innovation Systems, “are networks of institutions, public or private, whose activities and interactions initiate, import, modify, and diffuse new technologies.’” In ecological sociotechnical system work these innovation often disrupt existing models towards a greater good. Chiong Meza (2012) expanded on these ideals and they have many parallels in our study of open pedagogy. She suggested a System Agent Network approach where innovation systms lead to transitions over an extended period of time. Chiong Meza suggested transitions: * serve a need * transitions are complex phenomena * transitions occur through social desire * transitions happen at different speed. In an era of Artificial Intelligence we need data to foster open pedagogy as an innovation system that leads to positive transitions. We understand transforming education will take time and local context will determine the desire and speed of change. The united force is a set of learners who gather around a shared goal while making all of the tools and learning openly available. This in turn strengthens the system and encourages more agents to get involved. In the model there are three levels, the agent level, network level, and system level. System level events effect the network, and then the agent level. This causes the agents to seek types of technologies and interactions and they get involved and become actors. Often what drives action is a drastic occurrence at the system level that shocks agents to mobilize. After a drastic occurunc actors come to some general agreement of proposal to move forward and the actors begin to change their routines which over time influences the interaction mechanisms of the network. This mass has an influence of dynamics at the system level and positive change occurs overall. At the same time, Chiong Meza expanded System Agent model does not account for knowledge transfers among the three system levels. In our model of knowledge growth must of course be centered in any discussion of an innovation system. So we built on the SyNEA framework into the SAINT model: System Agent Intelligent Network Theory. The network itself must improve its corpus of knowledge. This idea has percolated for over a decade as we have watched the rise of Artificial Intelligence from when it was still called machine learning, because only the nerds could do it. The network of learning, a learner creates, through large language models acts as an Intelligent Agent. When we refer to Agent in this sense we do not necessarily mean an AI Agent, though those could very well be part of the system. For an average user, whether they are using a chat UI, or deployed an agent to read emails, the distinction does not matter. The machine is an agent that builds an intelligent network for the learner. We must capture how intelligence lives within both the people and the tools. Often the network itself acts as a teacher. ## Agentive Apprenticeship and the SAINT Model How does learning occur in innovation systems that can lead to positive transitions? One common element we think one must build into your innovation system means aligning someones personal learning goals with the values sought in the innovation. Thus we wanted to draw on the agentive value driven learning in our model. We don't want to wade into the "alignment" versus "safety" debate, but as classroom teachers we don't mind. Instead we build from past work in Cognitive Apprenticeship to and account that learning works best when the learner finds value in a positive environment. Therefore a theory of agentive apprenticeships begin with cognitive apprenticeship and build off of this original work. Cognitive Apprenticeship grew out of the work of Lagner and Wave and Communities of practice. In his work on Affinity Spaces, however, Gee noted that the growing networked nature meant that the geographic relationships do not fit. We build on this in Agentive Apprenticeship to account that the network and the space can act as an instructor as much as a person. ## Cognitive Apprenticeship Collins, Brown, and Duguid’s cognitive apprenticeship model argues that students learn complex thinking skills best by observing and practicing them in authentic contexts. Like a traditional apprenticeship, an expert makes normally hidden processes—such as problem-solving, interpretation, and self-monitoring—visible to learners. The model emphasizes six teaching practices: * Modeling: The teacher demonstrates and explains their thinking. * Coaching: Students practice while receiving guidance and feedback. * Scaffolding: The teacher provides temporary support. * Articulation: Students explain their reasoning. * Reflection: Students compare their work with expert or peer approaches. * Exploration: Support is gradually removed as students solve problems independently. Overall, cognitive apprenticeship treats learning as participation in the authentic practices of a community rather than simply acquiring isolated facts. Agentive apprenticeships just like Cognitive Apprenticeships requires knowledgeable instructors to impart knowledge to apprentices in a structured, “scaffolded” process (Brown, Collins, & Duguid, 1989). Scaffolding is defined as a series of instructional supports provided for the student during the learning process which is tailored to the needs of learners to allow them to achieve their learning goals (Sawyer, 2006). In apprenticeship learning target skills are not only continually in use by skilled practitioners, but are instrumental in completing meaningful tasks.Learning of skills and knowledge gets embedded in social and functional context. The biggest difference between agentive apprenticeships and cognitive apprenticeship is the focus on growing the network and spaces of learning rather than focusing solely on strategy use by the individual learner. ## Differences Between Cognitive Apprenticeship and Agentive Apprenticeship The two theories of cognitive and agentive apprenticeships have different design goals. cognitive apprenticeships are “designed to teach the process experts use to handle complex tasks."" An agentive apprenticeships we create networks designed to support community goals through learner growth In cognitive apprenticeship learning happens through guided experienced, while in agentive apprenticeship learning through a networked experience. Network Thinking in cognitive apprenticeships revolved around practices, norms, and enculturation. This requires extensive techniques to encourage the development of self correction and monitoring skills. In agentive apprenticeships these skills are encouraged, documented, and parsed using large language models, blogging and social media tools. They can use karma and gamification for documentation and dopamine, and should get driven by personal goal setting Network Technologies create an abstracted replay of learning. Yet this relies on a personal stewardship of data and data privacy. Agentive apprenticeships requires to think about identity as security and building boundaries around data we share with a network. Cognitive Apprenticeship worked through a strong focus on much about to deliver feedback. (see the technical reports from Brown, Collins, Dugiuid) and to encourage metacognitive stratgy use. Even here the metacognition, learning or thinking about your thinking and learning, was usually focused on outcome driven feedback. Agentive Apprenticeships create what James Gee labeled a "Circuit of Reflective Inquiry." Increasingly times this feedback will be provided by non human actors. Students already upload assignments and rubrics and ask AI for a grade. Their teacher may have even used AI to design the rubric. ## Sequencing in Agentive Apprenticeship Complexity becomes a key variable in agentive apprenticeships. When the delta between the man and the machine grows to great than the tools does not serve as a teacher. When creating a networked learning environment we can't let students slip far outside what Vygotsky called the "zone od proximal Development." We must also encourage diversity of thought and people in a network. Large Language Models will regress to the mean when it comes to divergent thinking. We need self programmable learners who can consider multiple perspectives and make evaluative judgement no machine has ever done. In terms of skills we must consider the global before local skills. When Palinscar and Brown developed reciprocal teaching they chose four cognitive skills they felt whern beyond reading. Fundamental and universal skills still matter. Text structure drives the corpus that feeds the machine. Encouraging networks to include print media strengthens the learner. Focusing on a global community while building universal math skills builds knowledge, ## Sociology of Learning in Agentive Apprenticeship When you read early debates of artificial intelligence folks always broke knowledge down into schemas or frames. Really this reflects a debate bewteen those who used information processing and the computer as a metaphor of the mind, and situativist who felt knowledge exists in action and environment as much as the mind. Agentive Apprenticeships situates itself within the canon of sysetms thinking and activity theory. From here we recognize that a culture of expert practice builds deep learning. Students need help in crafting networks that help to knit a community of knowledge/ Artificial Intelligence increases the need for intrinsic motivations for learning. Extrinsic pressures on learning, such as grades, points, and test scores can drive unethical or unknowing short cuts in students. Never begrudge a child for choosing the path of least resistance. We need to make the struggle effort worthy. We also need to encourage agents and people to promote and exploit Cooperation for their personal and network goals. LLMs usually work on very primitive reward systems and trying to correlate with human reviewers. Tom Sawyer gets his friends to Whitewash the fence. How different are the two scenarios? Networked learning means creating agentive apprenticeships that allow people to impact. In a networked society an agentive apprenticeships should also focus on Exploiting Competition. Signals get amplified with varying degrees of authenticity and authority. From tokens to town meetings networked society will get pushed by competing values and goals. img credit: "Apprentice" flickr photo by ManelTor shared under a Creative Commons (BY 2.0) license