A / compare_three_bots /exp_plan.md
linxinhua's picture
Session-recovery propagated from socratic: hidden gr.Textbox (#recovery-payload) + Gradio .then(js=...) chains write/clear localStorage; on_submit_initial / on_end_session yield JSON events on their CONFIRM yields (includes ENFORCE_END_SESSION_GATE toast path). RECOVERY_HEAD adds modal styling + window.__SOC_RECOVERY helpers; storage key socratic_frame_session_v1. on_recover_session restricts HF file lookup to this Space's mode_prefix (v5-HOT/socratic_frame/) because the dataset AE-Talk-bot/log-socratic is shared with socratic and a loose match would risk cross-condition recovery. TUTORIAL_HEAD walkthrough guard early-returns when localStorage carries an unfinished session. SessionLogger.from_dict restores dominant_framework + argument_submitted_at.
7edb5d1 verified
|
Raw
History Blame Contribute Delete
3.24 kB
- id: 1
component: claim
gap: well-addressed: ethically unacceptable
rationale: The draft clearly takes a side and argues the system should not be used. It frames the reason as failure to achieve desired outcomes rather than rights or consent.
anchor: I do not think this AI system should be deployed
- id: 2
component: stakeholders
gap: The draft names the university and implies students, but it does not clearly name which teachers or support staff will use the predictions.
rationale: A consequence argument needs a clear map of who is affected so the later outcome chain is concrete. Without naming teachers or other staff, it is easier for the argument to stay vague about who changes their behavior due to the system.
anchor: It seems that the university holds a main assumption
- id: 3
component: expected_consequences
gap: The draft argues the predictions cannot fully capture academic performance, but it does not lay out the effect chain from deployment to specific downstream harms or lost benefits.
rationale: Saying “not useful” does not yet show how decisions and behaviors will change after the system is used. A skeptical reader would ask what teachers do with the output, what students experience next, and what measurable learning outcomes follow.
anchor: students' academic performance cannot be entirely predicted
- id: 4
component: evaluation
gap: The draft does not weigh harms against benefits, it mostly states the system will not work as intended.
rationale: In a consequence-based argument, you need to compare the value of earlier intervention and suggestions against the value lost when the model misses how students learn. Without explicit weighing, the conclusion can feel asserted rather than judged.
anchor: I do not think this AI system is useful
- id: 5
component: alternatives
gap: Other options are not considered, aside from rejecting the system as not useful.
rationale: A reader may wonder whether a different design, different measures, or a narrower use could preserve the benefits. Without alternatives, the argument does not show why rejection is the best consequence-producing choice.
anchor: I do not think this AI system is useful
- id: 6
component: evidence_uncertainty
gap: The draft relies on an assumption about learning and predictability, but it does not say what evidence supports that assumption or what is uncertain.
rationale: Consequence arguments need a sense of what is known versus what is guessed. Without that, it is hard to tell whether the concern is a likely failure or only a theoretical one.
anchor: This is because students' self-regulated learning takes place in virtual contexts
- id: 7
component: final_judgement
gap: The strength of the “do not deploy” verdict is not calibrated to the level of certainty in the reasoning about predictive failure.
rationale: If the case does not show an error rate or evidence of poor outcomes, the conclusion needs to match that uncertainty. A skeptical grader might ask how strong the “not useful” claim is and what threshold of usefulness would change the decision.
anchor: I do not think this AI system is useful