{"actors":["Reference librarians and paraprofessional staff who provide in-person, chat, telephone, and email assistance","Public-services managers who schedule service coverage","Library systems analysts who maintain service-platform logs and derive workload measures","Patrons whose waiting time and access to appropriate assistance are affected"],"affected_objective":"Allocate reference-service coverage across times and channels while preserving timely access to staff attention and specialist escalation.","arm":"ORDINARY_DIVERSE_P2","authority_safety":{"authorized_first_step":"Conduct a retrospective, read-only analysis of de-identified service records from one completed scheduling period; keep derived workload trajectories outside scheduling and patron-record systems.","decision_authority":"The public-services manager retains staffing authority; service librarians adjudicate interaction and escalation boundaries, and the responsible privacy or data-governance officer determines whether records may be included.","excluded_actions":["No automatic change to staff schedules, assignments, performance evaluations, or service hours","No use of message content, patron identity, demographic attributes, or borrowing history","No treatment of unlogged activity as zero workload","No individual staff ranking or productivity scoring","No live routing, automated escalation, or denial of service during the first evidence step"],"halt_rollback":"Stop if patron or staff identities cannot be removed, timestamps cannot be reconciled across channels, concurrent activity cannot be separated from passive session time, or derived workload measures are used for individual evaluation. Delete the derived analytical dataset and leave the existing scheduling process unchanged."},"baseline":"Reference-service coverage is planned from weekly transaction totals and counts within fixed time and question-type bins. Each logged interaction contributes primarily as a count, so brief directional questions, extended consultations, overlapping chat sessions, abandoned waits, and specialist handoffs are represented through coarse categories or separate reports.","candidate_id":"discrete_continuous_model_selection__library_information_science__ORDINARY_DIVERSE_P2","causal_chain":["Transaction counts convert unequal amounts of staff attention into nominally equivalent units and divide workload at reporting-bin boundaries.","That false discreteness can make two periods with the same count appear equally demanding even when interaction duration, concurrency, and waiting burden differ.","Replacing counts with a single smoothed average would preserve aggregate effort but could erase short queue surges, abandonments, specialist handoffs, or channel outages that constrain coverage decisions.","A piecewise-continuous workload model represents active staff-attention demand, concurrency, and waiting burden as trajectories within staffed service intervals.","Explicit arrival, completion, abandonment, handoff, outage, and service-boundary events prevent interpolation across changes that reset or redirect those trajectories.","Cadence and boundary rules connect the modeled workload to shift-level allocation without treating every platform log entry as a separate unit of demand.","Retrospective comparison with transaction-count schedules and independent librarian review tests whether the representation changes coverage judgments for defensible reasons.","Only a separately authorized pilot may use a validated representation to inform future staffing decisions."],"cell_id":"discrete_continuous_model_selection__library_information_science","consequence":"Coverage decisions may place staff at times or channels that have many short transactions while leaving periods of sustained attention demand, overlapping conversations, waiting accumulation, or discrete service disruptions insufficiently visible.","diversity_from_prior_proposals":"This opportunity concerns cross-channel reference-service staffing and the conversion of interaction events into workload flow. Its affected problem, intervention, observable state, and causal path are separate from the sealed P1's digital-object preservation review and preservation-event trajectory.","experiment_id":"eoa_inverse_innovation_exp13_second_slot_policy60_20260806","intervention":"Construct a read-only, piecewise-continuous reference-workload model for each service channel. Within staffed intervals, estimate active staff-attention demand from de-identified interaction start and end times, documented active-handling intervals where available, concurrent conversations, and queue waiting time, sampled in five-minute windows. Do not infer zero demand from missing logs, and cap passive or unresolved sessions according to librarian-adjudicated rules. Retain arrivals, completions, abandonments, specialist handoffs, channel outages, and service openings or closures as discrete boundary events; do not interpolate through those boundaries. Compare alternative five-, fifteen-, and thirty-minute resolutions and revise the representation if channel technology, logging practice, service policy, or the scheduling decision changes.","mechanism_mapping":[{"counterfactual_removal":"Without a continuous process representation, unequal durations, concurrency, and accumulated waiting remain compressed into transaction counts.","mechanism_slug":"continuous_process_model","role":"Represents reference workload as changing staff-attention and waiting-demand trajectories rather than equally weighted transactions."},{"counterfactual_removal":"Without explicit boundary rules, smoothing could continue across abandonments, handoffs, outages, openings, or closures and misstate usable service capacity.","mechanism_slug":"hybrid_discrete_continuous_model","role":"Places discrete service events around piecewise-continuous workload intervals."},{"counterfactual_removal":"Without testing cadence, five-minute observations may create noise while longer windows may conceal short demand surges relevant to coverage.","mechanism_slug":"sampling_interval_choice","role":"Compares five-, fifteen-, and thirty-minute measurement windows against the scheduling decision."},{"counterfactual_removal":"Without a resolution audit, platform artifacts, passive sessions, and reporting-bin edges could be mistaken for real workload transitions.","mechanism_slug":"transition_resolution_audit","role":"Checks whether event definitions, session caps, and observation windows preserve reviewed changes in service demand."}],"nearest_rivals":["Weekly transaction totals by service channel","Fixed hourly transaction counts weighted by question category","Staff-reported busy-period summaries without reconstructed workload trajectories","A smooth average service-time forecast without explicit abandonment, handoff, or outage boundaries"],"negative_tests":{"intervention_falsifier":"The intervention is falsified if held-out scheduling periods show that workload trajectories do not yield more reproducible coverage judgments than transaction counts, if results reverse materially across reasonable cadence or passive-session rules, or if discrete boundary events add no decision-relevant information.","problem_falsifier":"The inferred problem is falsified if librarians make the same coverage allocations from transaction counts, duration-weighted counts, and piecewise-continuous trajectories, with no reviewed disagreements attributable to reporting bins, unequal attention demand, concurrency, waiting accumulation, or missed boundary events.","risks":["Log timestamps may measure platform connection rather than active staff attention.","Channels with richer instrumentation may appear more demanding than sparsely logged channels.","Five-minute trajectories may imply precision that the records do not support.","Long unresolved sessions may inflate estimated workload unless passive-time rules are explicit.","De-identification may be inadequate for rare specialist interactions or lightly used channels.","Derived workload measures could be repurposed for individual staff surveillance or performance evaluation.","A more detailed representation may increase analysis burden without altering coverage decisions."],"strongest_counterevidence":"If simple transaction counts or duration-weighted totals reproduce librarian coverage judgments across held-out periods, while concurrency and waiting measures are dominated by logging artifacts and boundary events do not alter staffing choices, the continuous workload representation is not justified."},"next_evidence_step":"Select one completed eight-week period from a single library service unit and use only existing de-identified timestamps and operational status records from at most three service channels. Two reference librarians independently mark a bounded sample of ambiguous handoffs, abandonments, passive sessions, and outages, then make coverage judgments for masked time blocks from transaction-count, duration-weighted, and piecewise-continuous representations. Compare judgment agreement, sensitivity to five-, fifteen-, and thirty-minute windows, errors around reviewed boundary events, and sensitivity to missing records. Do not alter schedules or contact patrons or staff represented in the records.","observable_state":"For each included service channel and staffed interval: de-identified interaction start and end timestamps; active-handling intervals when already recorded; queue-entry and service-start times; concurrent-session count; completion, abandonment, and specialist-handoff timestamps; channel availability and outage intervals; scheduled coverage; documented walk-up counts where available; and explicit indicators for missing or unreliable logging. Message content and personal identifiers are excluded.","prior_art_status":"UNSEARCHED","problem":"Reference-service staffing is planned from discrete transaction counts and fixed reporting bins even though the capacity burden is generated by continuously varying attention time, concurrency, and waiting accumulation, interrupted by consequential events such as abandonments, handoffs, and channel outages. The chosen granularity can therefore distort which periods and channels appear to require coverage.","proposal_index":2,"remaining_contrastive_claim":"For reference-service coverage, a piecewise-continuous workload representation is preferable to transaction counts or a fully smoothed forecast only if duration, concurrency, and waiting trajectories produce stable reviewed allocation differences while explicit event boundaries prevent consequential service changes from being averaged away.","revision_record":{"claim_changes":["Initial version states only a conditional comparison among workload representations and makes no claim about novelty, prevalence, demand, or effect size."],"conceptual_changes":["Initial candidate maps discrete–continuous model selection to reference-service staffing through continuous attention-demand trajectories bounded by service events."],"evidence_changes":["No external, repository, or prior-art evidence was inspected; prior art remains unsearched."],"operational_changes":["Initial evidence step is limited to a read-only eight-week retrospective analysis of at most three channels in one service unit, with no scheduling changes."],"parent_version":null,"progress_targets_addressed":["Independent domain problem and affected objective","Explicit false-discreteness and false-smoothness costs","Observable state, granularity, cadence, and boundary rules","Mechanism and structural mappings","Authority limits, privacy safeguards, falsifiers, risks, and bounded evidence step"]},"schema_version":1,"structural_mapping":[{"archetype_element":"Decision Need","domain_realization":"Allocate reference-service coverage across times and channels under finite staff availability."},{"archetype_element":"Process Change Signature","domain_realization":"Staff-attention demand, concurrency, and waiting burden vary over time, while arrivals, completions, abandonments, handoffs, outages, openings, and closures occur as discrete events."},{"archetype_element":"Cost of False Smoothness","domain_realization":"Averaging across short surges, abandonments, handoffs, or outages can hide periods when coverage constraints change abruptly."},{"archetype_element":"Cost of False Discreteness","domain_realization":"Equal transaction counts and fixed reporting bins fragment sustained work, erase duration and concurrency differences, and create artificial edges between adjacent periods."},{"archetype_element":"Granularity Choice","domain_realization":"Use channel-specific piecewise-continuous workload trajectories within staffed intervals, initially observed in five-minute windows, with native timestamps for boundary events."},{"archetype_element":"Step Boundary","domain_realization":"A workload segment starts or ends at a service opening, closure, outage, recorded handoff, or other adjudicated event that changes available capacity or responsibility; reporting-bin edges alone do not create transitions."},{"archetype_element":"Continuity Assumption","domain_realization":"Attention and waiting demand may be interpolated only within a staffed, available channel interval when no boundary event or logging gap intervenes."},{"archetype_element":"Measurement Resolution","domain_realization":"Five-minute workload estimates are compared with fifteen- and thirty-minute alternatives, while original event timestamps and missingness indicators are retained."},{"archetype_element":"Hybrid Boundary Rule","domain_realization":"Continuous workload estimation operates between boundary events; an abandonment, specialist handoff, outage, opening, or closure terminates interpolation and begins a separately defined segment."},{"archetype_element":"Transition Validation","domain_realization":"Independent librarians review ambiguous events and compare coverage judgments across count-based, duration-weighted, and continuous representations."},{"archetype_element":"Approximation Error Check","domain_realization":"Assess allocation disagreements caused by equal-event weighting, bin edges, excessive smoothing, passive-session assumptions, sparse logging, and cadence choice."},{"archetype_element":"Scale Shift Review","domain_realization":"Reassess the representation before applying it to a service unit with different channels, logging systems, staffing intervals, or scheduling decisions."}],"title":"Piecewise-Continuous Workload Modeling for Reference-Service Coverage","version":0}