Campfire Session — Customizing your assistant: Sparky!

Introduction • 00:00

• Lulu introduces the session and agenda.

Ice-breaking news • 1:29

• A discussion explores Howard Gardner’s vision for the future of education and intelligence in the age of AI, as shared in a recent Harvard Gazette article. Gardner predicts that by 2050, students may require only a few foundational years of schooling—focused on reading, writing, arithmetic, and basic coding—before shifting toward self-directed learning guided by teachers acting as coaches and mentors.

• Emphasis is placed on the evolving role of teachers and the growing potential of AI to enable truly personalized learning, where education adapts to each student’s curiosity, strengths, and aspirations rather than adhering to a uniform, time-bound model of schooling.

How Sparky memories work • 4:12

• Memories can be set for Sparky during chats and activities, and Sparky can remember preferences and teaching context.

• Sparky can differentiate between teacher and student roles, acting as a more helpful teacher and tutor to students.

• Users can specify memory details like who they are, what grade they teach, and preferred teaching style; these memories persist across chats and activities.

Configuring Sparky: settings overview • 10:20

• Sparky’s build is shown manually, detailing five sections and how to customize its behavior.

• The overview covers activity naming, summaries, and then emphasis on behavior, including persona, goals, and interaction style.

• The discussion outlines grade level and language proficiency settings, plus a role/policy for Sparky’s persona and response style.

• Guidelines for Sparky's responses, greeting messages, and optional response suggestions for students are introduced.

• Sparky supports multiple languages with different dialects and can handle secondary language scenarios.

• There are about 500 languages available; Sparky can understand and respond in those languages, with some limitations on entertaining non-primary languages.

• Sparky offers text-to-speech and speech-to-text capabilities; students can speak aloud to Sparky and Sparky can speak aloud to students.

• The system allows various interaction forms, including removing writing, having students listen, or combining features.

• Grading rubrics can be auto-graded or not, can be shown to students, and can use different scales (A-F, 0-100, bucketed into 10-point or 1-7 etc.)

• Assignments have housekeeping controls to pause/unpause activities, set durations, and manage access, including release dates, deadlines, and the ability to remove or re-submit sessions.17:06

• Session control includes pausing and unpausing to allow continuation or new sessions, with insights into when students should finish.

• Voice customization is not yet available; Sparky uses a fixed library of voices per language, and deeper or higher-pitched tones may occur contextually.

Workspace and group settings • 19:54

• Lulu covers group setting, like matching grade levels to goals, and handling copying assignments between groups with potential overrides.

• Discussion confirms top-level rules apply with priority; testing suggested to ensure settings propagate correctly across groups.

• Sparky features are explained, including support for mixed-language classes, subgrouping, and adaptive input based on proficiency.

• Moderation and mission/background settings are discussed, including character limits, etiquette, academic integrity, and localizing content to reduce Western centric bias.

• Practical guidance on configuring Sparky for diverse classrooms, plus templates, memory settings, and workspace/moderation options are covered.

Shareout and feedback • 32:32

• Colin Bridgewater shares how Flint’s Sparky features have been used to tailor mission statements and school guidelines, including character limits and region-specific contact information.

• Colin explains practical enhancements: submission reminders, teacher scripts, and eventual adoption of Sparky across learning activities while noting improvements in feature availability over three months.

• Kat Gorringe explains memories feature, sharing steps to save DBQ configurations as core memories, and demonstrates long-term consistency in formatting across DBQ outputs.

• Kelly Schuster-Paredes expresses appreciation and acknowledges broader Flint community adoption while acknowledging established connections with Sarah Ducharme.

Conclusion • 47:43

• Lulu outlines future Flint engagement opportunities. The platform continues to expand its educational support ecosystem.

• Lulu also shares QR codes for people to check out the Campfire Calendar, Flint's Instagram (which has a bunch of teacher-facing content), and the Flint Community.