Periodic Reporting for period 2 - CO-ADAPT (CO-ADAPT: Adaptive Environments and Conversational Agent Based approaches for Healthy Ageing and Work Ability)
Reporting period: 2019-12-01 to 2021-05-31
OBJ 1. Propose a comprehensive CO-ADAPT framework to support change for healthy ageing that describes types of adaptations required in work environments and how to approach adaptation of ageing citizens in the form of change behaviour and learning ICT. The framework is produced as a participatory effort including relevant stakeholder representatives including prospective ageing users.
OBJ 2. Create a personalised conversational application with comprehensive change support for ageing citizens based on data analytics focusing on high-level strategies and latest change psychology approaches considering wellbeing and digitalisation across work and personal life. Studying cross cultural applicability by evaluating the CO-ADAPT application and adaptations in work systems in two different cultural contexts in Italy and Finland.
OBJ 3. Develop secure and GDPR compliant AI techniques capable of contextually recommending change programs.
OBJ 4. Investigate three types of smart adaptations in work systems to age thresholds, to individual capabilities, information adaptations to tasks for cognitive augmentation aiming at 2 digits percent improvements on most indexes.
Main achievements concern the consolidated CO-ADAPT framework. We have motivated subprojects with the conceptual framework founded on workability and wellbeing, we clarified the links between subprojects through exploratory Synergy studies with research questions.
The users' concerns about privacy and ethical issues in the usage of adaptive systems are reviewed in the scientific literature. A set of recommendations are derived to inform the design and development of adaptive technologies. A set of interviews and a remote study have been completed with aging users to test the usability of an interface that features conversational agent interaction and behavioral change goals.
The implementation generalizability of the smart shift scheduling was finalized according to the plan. The Adaptive Assembly Workstation (AAW) equipped with a cobot is implemented that is capable of adapting to the operator's internal state during the work cycle. Entity recommender system was reimplemented; the new implementation includes a functionality that allows users to exclude applications or software from monitoring and the new user interface follows elderly-friendly design.
The CO-ADAPT privacy preserving platform has been delivered. The participants’ data have been collected at the study sites, aggregated into a JSON object and encrypted for secure transfer to the platform. At the platform side the data are decrypted and split into the constituent psychological and physiological entities, to be stored in the platform and made available for analytics visualisations.
The analytics portal app offers a front-end to the platform for managing the organisational and user entities involved in studies, as well as for offering the analytics visualisations.
The CO-ADAPT conversational system has been developed and incrementally tested in three experimental rounds. The participants received eight sessions of stress management counselling by licensed psychotherapists, and the CO-ADAPT conversational system provided them with support and suggestions for applying the stress reduction techniques learnt during the sessions with the counsellors.
Knowledge base for behavioral change tips for healthy lifestyle was implemented and released as open source.
A remote study with 176 shift workers between ages 40 and 71 (median 50) has been conducted to evaluate the utility of different question prompts during behavioral recommendation interaction.
A predictive model has been trained on the data to assess the predictive power and behavioral effects of different question prompts. The analysis of the results are ongoing. All the results from the first and second round of the experimental data collections have been analysed and statistically evaluated.
We have been creating content regarding project activities to grow our audience in the several dissemination channels. We have a steady growth of followers. The project partners participated (with presentations, poster presentations or for networking reasons) in several events. We have also established important synergies and collaborations with related projects and initiatives that resulted in two reports.
Based on our studies, the use of the smart shift scheduling tool decreased sickness absence by nearly 10% and showed beneficial well-being effects on especially perceived work ability compared to employees remaining in traditional scheduling. It is a promising tool to support shift work management for the reduction of sickness absences and promoting well-being among hospital employees. The conservative cost-benefit estimates from the employer point of view show that participatory scheduling´s benefits outweigh the costs.
The adoption of collaborative adaptive robotics aims at preserving the aged workforce insofar as the cobot are not a replacement of senior workers. Besides, the embedded algorithm is conceived to reduce the effects of the determinants of errors, low satisfaction, and productivity on working life, and this is expected to produce beneficial spill-over effects on the non-working life. Finally, a decrement in determinant stress/mental overload will result in a reduction in safety issues, errors, and as a consequence in work-related injuries and health-related costs.
Proactive Entity rec
Participatory design session for entity recommender system suggests both young and older adults appreciate the support of the recommender in a variety of situations. Results from real-world study suggest that proactive entity recommender was found useful and obtained large improvements in the user’s task performance and the improvements were significant. In order to validate the effectiveness of entity recommendations on older adults, we reimplemented the system following the guidelines for elderly-friendly design. Results from pilot study indicate that older adults used and benefited from the recommended entities. The actual study with 2x2 factorial design (age group, with/without recommendation) (N=30) is being conducted with the aim to examine whether older adults could find recommendations more useful, compared to the younger adults.
Conversational Agent CO-ADAPT Application
The different versions of the Conversational Agent, increasingly able to establish a dialogue with the users, have been designed, developed and tested with real ageing users and psychotherapists. With respects to currently available applications for mental wellbeing, UNITN’s CO-ADAPT Conversational Agent goes beyond state-of-the-art because it is designed to in terms of natural language understanding of users’ input, training to engage in personalized dialogues and across a long period of time; the component technologies underlying these capabilities are based on machine learning algorithms which have been trained with data collected in the three rounds of data collection.