The Science in Shiftic
Summary
Shiftic is a research-informed workspace for designing and running behavior change initiatives. The science behind Shiftic informs different parts of the product: how you set goals for interventions, how the journey is shaped, how activities are scored, and how you follow up on results. This page is a guided tour of where that science lives and where it comes from.
How it works: the map
Shiftic is built in two layers.
Underneath sits a general-purpose language model — the same kind of AI that powers many AI tools today. It's good at language, explanation, and generation. On its own, it doesn't know your domain, your plan, or your target group.
On top sits Shiftic's own features, frameworks, and workflows — the Behavior Shift Score® (BSS), the way Shiftic helps structure briefs and interventions, the Impact Loop, and more. These are what give the AI its specific behavioral-science lens. Each feature is built on a specific body of research.
Different parts of Shiftic use different science:
When you plan interventions — setting the goal for a specific intervention. Shiftic helps classify what kind of change the intervention aims for: a capability (people can do X), an opportunity (conditions exist to do X), or a motivation (people are inclined to do X). This is based on the COM-B model of behavior change (Michie et al., 2011).
When you design the journey of interventions. Shiftic's AI draws on established change frameworks to shape phases, pacing, and tactics - based on the context and content of each unique project you create. Examples include Kotter's 8-Step model, ADKAR, Lewin, Nudge Theory, and Kübler-Ross's change curve. These shape what Shiftic suggests — they do not contribute to the Behavior Shift Score.
When you design activities and check plan strength with the Behavior Shift Score. The Behavior Shift Score® (BSS) is Shiftic's core scoring tool. It checks whether your plan pulls the most effective behavioral levers by analyzing your activities against a framework of 24 behavioral determinants. 16 of these come from the 2024 Albarracín review in Nature Reviews Psychology; 8 additional ones were developed with Research Institutes of Sweden(RISE) for organizational contexts.
When you follow up on the impact of an intervention(Impact Loop). Shiftic uses a proof-signal approach to help you distinguish actual behavioral evidence from self-reported satisfaction — capturing what you observed or relevant data you have access to, not what participants reported feeling. The COM-B model (again — see the intervention goals bullet above) informs how proof signals are structured for different kinds of interventions.
A practical way to think about it:
Some things you see in Shiftic are produced by the research-backed frameworks directly. Others are the AI generating language, ideas, and suggestions on top. Both are useful, but they're not the same kind of output.
The Behavior Shift Score itself (the number, and the observation that your plan is strong in some areas and weak in others) comes from Shiftic's framework. If your plan is light on structural factors, the score reflects that because the research says structural factors matter.
The specific ideas and suggestions written in the chat — "you could add a peer coaching circle," "try a manager check-in every other week" — come from the AI. The framework may have pointed out that you need more social support; the specific method is the AI filling in the blank using general knowledge and your context.
These two kinds of output feel similar in the chat, but they have different backing. The framework-level observations are tied to research. The specific method suggestions are the AI being creative within the framework — they're useful starting points, but your domain expertise decides if they actually fit your situation.
Where does all the research actually live? The research was used when we built these frameworks — to decide which 24 determinants matter, how to rank them, what COM-B distinguishes, which change models to draw on. But the research papers themselves aren't stored inside Shiftic. The product runs on the frameworks we built from the research, not on a searchable library of the research itself. That's why, if you ask the chat "what paper supports this?", the answer can be unreliable — it doesn't have the papers to look up (more on this below).
The tour: going deeper on each area
Intervention goals: what kind of change does this intervention aim for?
This is about setting the goal for a specific intervention, not the overall vision of your brief.
One of the most common design mistakes in behavior change is treating all interventions as if they target the same kind of change. An intervention aimed at building a skill needs a different design than one aimed at shifting confidence or one aimed at removing a logistical barrier. If you don't know which kind of change you're after, you can easily end up with interventions that look right on paper but don't match the actual barrier.
Shiftic helps by classifying intervention goals using the COM-B model — a behavior-change framework that distinguishes:
Capability — the person can do the behavior (knowledge, skills, confidence)
Opportunity — the conditions let the person do it (time, tools, access, social context)
Motivation — the person is inclined to do it (emotional, reflective)
Examples of goals for the same overall initiative, by type:
Capability goal: "The team can run a retrospective on their own without external facilitation."
Opportunity goal: "The team has time blocked on the calendar and a shared template to run retros."
Motivation goal: "The team feels the effort of running retros is worth it."
Each type calls for a different kind of intervention. Shiftic uses this classification to orient your design — it doesn't decide for you, but it helps make the distinction visible.
Designing the journey of interventions
Behavior change is a journey, not an event. Awareness, capability, and reinforcement unfold across phases. A sequence of well-placed interventions does something that a single workshop can't.
Shiftic's AI, depending on the nature of your initiative, may draw on established change frameworks to shape how the journey is structured. A few examples:
Kotter's 8-Step model — for large strategic transformations with leadership alignment and cultural anchoring.
Prosci ADKAR — for individual capability-building journeys (Awareness, Desire, Knowledge, Ability, Reinforcement).
Lewin's Unfreeze-Change-Refreeze — for shifting entrenched patterns.
Thaler & Sunstein's Nudge — for choice-architecture and small behavioral cues.
Kübler-Ross's change curve — for anticipating the emotional phases people go through in response to change.
These frameworks shape what the chat suggests when designing your journey. They do not contribute to the Behavior Shift Score. The score is calculated separately, against the 24-determinant framework (see next section).
Designing activities: are you pulling the right levers? (the Behavior Shift Score)
The Albarracín review identified a striking pattern across behavioral domains: different types of interventions have vastly different effects. Access-type interventions (removing friction, making the behavior easy) are about 5× more effective than information-type interventions (telling people why the behavior matters). Yet most organizational plans lean heavily on information and awareness. That's a costly blind spot.
Shiftic's Behavior Shift Score® (BSS) checks whether your plan is actually pulling high-leverage factors. It analyzes your activities against a framework of 24 behavioral determinants and produces a relative design-strength signal.
BSS is:
A design-quality check, before you launch
A way to catch blind spots (over-reliance on low-impact levers)
A shared language for comparing plan variants
BSS is not:
A prediction of outcomes
A measure of actual effect
A verdict on whether an activity is "good" (many activities matter for reasons the score doesn't capture — communication, legal, cultural)
How BSS works at a high level: Shiftic classifies each activity against the 24 determinants, applies research-based weighting, balances across activities and the length of the initiative, and produces a score. Exact weights and formulas are part of Shiftic's proprietary model.
The 24 determinants. 16 come from the Albarracín review; 8 are additions developed with RISE for organizational change contexts. A few examples to illustrate the concept — not a full list.
A few examples of individual factors (within the person):
Habits (medium effect) — automatic routines that run without conscious effort. Once established, they keep shaping behavior even when motivation dips.
Emotions (small) — feelings like hope, pride, or fear that pull people toward or away from an action.
Beliefs (minimal) — what people think will happen if they do something. Useful to understand, but changing beliefs alone rarely changes behavior.
Knowledge (minimal) — knowing the facts about a behavior. Necessary for awareness, but on its own has almost no effect on what people actually do.
A few examples of social-structural factors (around the person):
Access (large) — whether it's physically, logistically, or financially easy to do the behavior. Removing friction consistently has the biggest effect across domains.
Social support (medium) — people around you who help, remind, or encourage the behavior.
Material incentives (small) — tangible rewards. Works to start change, fades without deeper reinforcement.
Legal & administrative sanctions (minimal) — rules and penalties that compel the behavior. Can produce short-term compliance but rarely sustained change on their own.
A few examples of what RISE added for organizational contexts:
Active Practice (large) — structured, hands-on opportunities to try the behavior in real situations, with feedback and adjustment. The opposite of one-off training.
Leadership Influence (medium) — what leaders visibly do, model, and reinforce. When leaders practice and prioritize the behavior, adoption spreads faster.
Seamless Integration (large) — building the behavior into everyday work and tools, so it happens in the flow of normal activity instead of as a separate effort.
Instant Feedback (medium) — immediate, specific input on how someone is doing, so they can adjust in real time.
Reading the score: BSS is a relative design strength signal. A score in the 30–50 range signals reasonably strong behavioral design. Higher scores reflect more powerful combinations of interventions; lower scores may indicate missing ingredients or over-reliance on low-impact activities. The numbers may look like percentages, but they aren't tied to real-world baselines yet — the current model uses a placeholder starting assumption, to be replaced with empirically grounded baselines in future versions.
Important qualifier about the RISE extensions. The OR values for the 8 RISE-added determinants come from correlational and theoretical research, not from intervention meta-analyses like the Albarracín ORs. They are informed estimates for organizational contexts, not intervention-validated effect sizes on par with the Albarracín ones.
Following up: did it actually move? (the Impact Loop)
A plan design is a hypothesis. The Impact Loop in Shiftic helps you test that hypothesis by guiding you to capture data and what you observed after activities run.
The research is clear: self-report ("I feel more confident") is a weak indicator of behavior change. Observable behavioral evidence ("I saw the team raise this topic unprompted three times this week") is much stronger.
Shiftic uses a proof-signal approach: you describe, ahead of time, what specific, observable indicator would tell you the behavior is actually happening. Then you evaluate whether you saw it.
Example:
Weak signal: "Participants report feeling confident in the new process."
Proof signal: "At least 3 team members raise the new process unprompted in the next team meeting."
The COM-B distinction (capability, opportunity, motivation) helps structure proof signals so they fit the type of change the intervention was aiming for. A capability intervention has different observable signals than a motivation intervention.
The detailed internal methodology here is more proprietary than the other three areas and has fewer external citations.
About Sources and Citations in Chat
When you ask Shiftic in chat "what research supports this?", the AI will attempt to provide references — but these come from the underlying language model's general training data, not from a database of Shiftic's actual sources. As the map above explains: the research behind Shiftic was used to build the frameworks, not stored as a searchable knowledge base inside the product.
This means citations in chat can sometimes be irrelevant, out of context, or even fabricated. The trustworthy view of Shiftic's research foundation is this help page. We're actively working on making real sources accessible to the AI so it can cite them correctly.
Recommended for regulated contexts: if you need to anchor decisions in specific research, use this help page as the authoritative source. Avoid using Shiftic to generate bibliographies or literature lists — use it as a design aid alongside your own domain research.
Working with Your Own Context
The science behind Shiftic is general and cross-domain by design. To make it relevant for your specific context:
Upload your own material into the project — strategy documents, prior research, internal methods, domain-specific guidelines, target group insights. Shiftic uses this material as active context when supporting your planning.
Bring your domain expertise. The frameworks tell you which types of levers tend to matter. What the right implementation looks like for your specific population, culture, or setting — that's your expertise.
Combine. In practice you're pairing the general behavioral science model with your own domain knowledge, which gives context-relevant AI support rather than generic output.
When Shiftic Helps (and When It Doesn't)
Shiftic is most useful when you want a structured way to think through a change initiative. It doesn't tell you what to do — it shows you the building blocks research has linked to effective behavior change, so you can check whether your plan considers them. You decide what fits your context.
Use it when:
You're delivering an initiative with the goal for people to behave differently (new ways of working, new tools, cultural change, onboarding, competence development) and want to think through whether you've covered the factors that matter.
You want a structured way to check plan quality before investing in implementation.
You want a shared language to compare plan variants or iterations with your team.
You want to catch blind spots — for example, a plan heavy on information but light on structural support.
Not ideal for:
Operational tasks without a behavior change dimension.
Projects that are purely technical or administrative.
Contexts where you need specific literature references or formal scientific grounding (see "About Sources and Citations in Chat" above).
Situations where you need certified validation against specific legal requirements, regulatory frameworks or industry standards.
Rule of thumb: if the initiative requires someone to do something differently, Shiftic is likely relevant. If not, it probably isn't.
Current Limitations
This is an Alpha version of the BSS model — useful, but intentionally simplified. Known limitations we're actively working to address:
Prototype stage: the model is not final and will improve as it evolves.
No causation: the model identifies correlations between determinants and behavior change, not direct causes.
No dose modeling: a 10-minute e-learning and a two-day workshop are treated equally if they target the same determinant.
No timing or sequencing: activities are scored statically, without accounting for when or how they're delivered.
No interaction effects: interventions are assumed to add up independently; in reality, they may amplify or dampen each other.
No audience variation: the model doesn't yet adjust for differences across participant groups.
No source traceability in chat: the AI doesn't have access to the actual research sources. Citations provided in chat may be inaccurate and should be verified independently.
Framework/AI boundary not yet visible in the UI: users currently can't easily distinguish framework-level insights (research-backed) from AI-generated implementation suggestions.
A Note on Correlations vs. Causation
At this stage, the BSS model focuses on identifying correlations between determinants and behaviors — not establishing direct causation. Correlations indicate a relationship between variables but do not confirm that one causes the other.
This is a deliberate and necessary first step. By mapping these connections now, Shiftic is building the foundation for deeper causal studies and more actionable predictions in the future.
Frequently Asked Questions
Is there one scientific framework Shiftic defaults to? No. Different parts of Shiftic are built from different research traditions, as shown in the map and the tour above. COM-B informs goal-setting for interventions and proof-signal structuring; the 24-determinant framework informs the Behavior Shift Score; various change frameworks inform how Shiftic shapes journeys.
If the research isn't stored in the product, what is? The frameworks built from that research — the 24 determinants, the COM-B classification, the change-model patterns the AI can draw on, the proof-signal methodology. The research papers themselves were used at design time as input. They aren't queryable by the AI at runtime.
What does BSS measure? How well your plan design matches the factors research shows drive behavior change. A design-phase quality check, not a success prediction.
What is BSS not? Not a measure of actual effect. A high score doesn't guarantee success; a low score doesn't disqualify an initiative. It's one perspective among several.
An activity got a low score — should we drop it? Not necessarily. The score reflects behavior-change contribution per the model. An activity can still be important for anchoring, communication, legal, or cultural reasons. Use the score as one input, not a verdict.
How can we improve our score? Shiftic gives concrete suggestions: complement information with practical application, add peer learning, strengthen structural enablers. You decide what fits your context and constraints.
We work in a regulated context (need certified validation against specific legal requirements, regulatory frameworks or industry standards) — how do we handle the research backing? The underlying model is grounded in solid research, but the platform does not currently make it easy to trace decisions back to original sources. Our recommendation is to familiarize your team with the 24 behavior factors and the Albarracín review as complements to using the platform. Shiftic supports planning — it doesn't replace your professional judgment or responsibility for grounding decisions in your specific requirements.
The research is general — our context is specific. How is that handled? Upload your own material (organization-specific methods, domain research, prior initiatives, internal models) directly into the project. Shiftic reads and uses it as active context. In practice you're combining the general behavioral science model with your own domain expertise.
Why did Shiftic cite a paper that doesn't fit my context? The AI doesn't have access to Shiftic's actual research sources in chat. When asked for references it falls back on its general training data, which can surface papers that are off-target or, in some cases, fabricate references. Treat chat citations with caution; this help page is the trustworthy view of the research foundation.
References
The references are structured by which part of the product they support. This mirrors the map above — so you can see, for any area of Shiftic, which science informs it.
A. Sources behind the 24-determinant framework (used by the Behavior Shift Score)
Primary foundation:
Albarracín, D., Fayaz-Farkhad, B., & Granados Samayoa, J. A. (2024). Determinants of behaviour and their efficacy as targets of behavioural change interventions. Nature Reviews Psychology, 3(6), 377–392. https://doi.org/10.1038/s44159-024-00305-0
This review is itself a synthesis of meta-analyses across behavioral domains. The 16 original determinants, their relative impact rankings, and the OR interpretation scale used in BSS all come from this review.
Sources for the 8 RISE-added determinants:
Deci, E. L., & Ryan, R. M. (2000). The "what" and "why" of goal pursuits. Psychological Inquiry, 11(4), 227–268.
Gagné, M., & Deci, E. L. (2005). Self-determination theory and work motivation. Journal of Organizational Behavior, 26(4), 331–362.
Ericsson, K. A., Krampe, R. T., & Tesch-Römer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100(3), 363–406.
Salas, E., Tannenbaum, S. I., Kraiger, K., & Smith-Jentsch, K. A. (2012). The science of training and development in organizations. Psychological Science in the Public Interest, 13(2), 74–101.
Cialdini, R. B., & Goldstein, N. J. (2004). Social influence: Compliance and conformity. Annual Review of Psychology, 55, 591–621.
Judge, T. A., & Piccolo, R. F. (2004). Transformational and transactional leadership. Journal of Applied Psychology, 89(5), 755–768.
Schein, E. H. (2010). Organizational Culture and Leadership. Jossey-Bass.
Denison, D. R. (1990). Corporate Culture and Organizational Effectiveness. Wiley.
Bass, B. M. (1990). Bass & Stogdill's Handbook of Leadership: Theory, Research, and Managerial Applications. Free Press.
Yukl, G. (2013). Leadership in Organizations (8th ed.). Pearson.
Kluger, A. N., & DeNisi, A. (1996). The effects of feedback interventions on performance. Psychological Bulletin, 119(2), 254–284.
Hattie, J., & Timperley, H. (2007). The power of feedback. Review of Educational Research, 77(1), 81–112.
American Society for Training and Development. (2010). The Accountability Study on Goal Setting and Follow-up.
B. Source behind the COM-B classification (used for intervention goals and proof-signal design)
Michie, S., van Stralen, M. M., & West, R. (2011). The behaviour change wheel: A new method for characterising and designing behaviour change interventions. Implementation Science, 6, 42.
C. Frameworks Shiftic's AI draws on for journey design
These frameworks shape what Shiftic suggests when designing the journey of interventions. They do not contribute to the Behavior Shift Score calculation.
Kotter, J. P. — 8-Step Change Model (large strategic transformations)
Lewin, K. — Unfreeze-Change-Refreeze (shifting old patterns)
Prosci ADKAR — Awareness, Desire, Knowledge, Ability, Reinforcement
McKinsey — 7S Framework (system-wide alignment)
Thaler, R., & Sunstein, C. (2008). Nudge (behavior change through choice architecture)
KAIROS Change Model (emergent/adaptive change)
Bloom, B. S. — Taxonomy of learning objectives
Theory of Change / Impact Chain (mapping interventions to outcomes)
Moore, C. (2017). Action Mapping (behavior-focused training design)
Kübler-Ross, E. — Change Curve (emotional response phases)
BCG Henderson Institute — AI-enabled learning / 90/10 pattern research
D. Impact Loop methodology
Largely an internal methodology, drawing conceptually on the COM-B distinction for structuring proof signals by change type. No separate external citations currently listed.
Disclaimer
Shiftic's scoring and design tools are directional design indicators, not guarantees of real-world outcomes. As the product is continuously developed and applied using AI, occasional inconsistencies or errors may occur. The Behavior Shift Score does not process personal data and is provided "as is" — the user remains solely responsible for decisions made based on its outputs.
© Shiftic AB, 2026. All rights reserved. The Behavior Shift Score® is a registered trademark and intellectual property of Shiftic AB and may not be copied, shared, or used without prior written permission.