skip to main content

Title: Developing common qualitative tools for cross ERC education program evaluation
National Science Foundation (NSF) funded Engineering Research Centers (ERC) are required to develop and implement education and outreach opportunities related to their core technical research topics to broaden participation in engineering and create partnerships between industry and academia. Additionally, ERCs must include an independent evaluation of their education and outreach programming to assess their performance and impacts. To date, each ERC’s evaluation team designs its instruments/tools and protocols for evaluation, resulting in idiosyncratic and redundant efforts. Nonetheless, there is much overlap among the evaluation topics, concepts, and practices, suggesting that the ERC evaluation and assessment community might benefit from having a common set of instruments and protocols. ERCs’ efforts could then be better spent developing more specific, sophisticated, and time-intensive evaluation tools to deepen and enrich the overall ERC evaluation efforts. The implementation of such a suite of instruments would further allow each ERC to compare its efforts to those across other ERCs as one data point for assessing its effectiveness and informing its improvement efforts. Members of a multi-ERC collaborative team, funded by the NSF, have been leading a project developing a suite of common instruments and protocols which contains both quantitative and qualitative tools. This paper reports on the development of a set of qualitative instruments that, to date, includes the following: (a) a set of interview/focus group protocols intended for various groups of ERC personnel, centered around five common topics/areas, and (b) rubrics for summer program participants' verbal poster/presentations and their written poster/slide deck presentation artifacts. The development process is described sequentially, beginning with a review of relevant literature and existing instruments, followed by the creation of an initial set of interview questions and rubric criteria. The initial versions of the tools were then pilot-tested with multiple ERCs. Feedback sessions with education/evaluation leaders of those piloting ERCs were then conducted, through which further revision efforts were made.  more » « less
Award ID(s):
Author(s) / Creator(s):
; ; ;
Date Published:
Journal Name:
American Society for Engineering Education
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. National Science Foundation (NSF) funded Engineering Research Centers (ERC) must complement their technical research with various education and outreach opportunities to: 1) improve and promote engineering education, both within the center and to the local community; 2) encourage and include the underrepresented populations to participate in Engineering activities; and 3) advocate communication and collaboration between industry and academia. ERCs ought to perform an adequate evaluation of their educational and outreach programs to ensure that beneficial goals are met. Each ERC has complete autonomy in conducting and reporting such evaluation. Evaluation tools used by individual ERCs are quite similar, but each ERC has designed their evaluation processes in isolation, including evaluation tools such as survey instruments, interview protocols, focus group protocols, and/or observation protocols. These isolated efforts resulted in redundant resources spent and lacking outcome comparability across ERCs. Leaders from three different ERCs led and initiated a collaborative effort to address the above issue by building a suite of common evaluation instruments that all current and future ERCs can use. This leading group consists of education directors and external evaluators from all three partners ERCs and engineering education researchers, who have worked together for two years. The project intends to address the four ERC program clusters: Broadening Participation in Engineering, Centers and Networks, Engineering Education, and Engineering Workforce Development. The instruments developed will pay attention to culture of inclusion, outreach activities, mentoring experience, and sustained interest in engineering. The project will deliver best practices in education program evaluation, which will not only support existing ERCs, but will also serve as immediate tools for brand new ERCs and similar large-scale research centers. Expanding the research beyond TEEC and sharing the developed instruments with NSF as well as other ERCs will also promote and encourage continual cross-ERC collaboration and research. Further, the joint evaluation will increase the evaluation consistency across all ERC education programs. Embedded instrumental feedback loops will lead to continual improvement to ERC education performance and support the growth of an inclusive and innovative engineering workforce. Four major deliveries are planned. First, develop a common quantitative assessment instrument, named Multi-ERC Instrument Inventory (MERCII). Second, develop a set of qualitative instruments to complement MERCII. Third, create a web-based evaluation platform for MERCII. Fourth, update the NSF ERC education program evaluation best practice manual. These deliveries together will become part of and supplemented by an ERC evaluator toolbox. This project strives to significantly impact how ERCs evaluate their educational and outreach programs. Single ERC based studies lack the sample size to truly test the validity of any evaluation instruments or measures. A common suite of instruments across ERCs would provide an opportunity for a large scale assessment study. The online platform will further provide an easy-to-use tool for all ERCs to facilitate evaluation, share data, and reporting impacts. 
    more » « less
  2. The Engineering Research Centers (ERCs), funded by the National Science Foundation (NSF), play an important role in improving engineering education, bridging engineering academia and broad communities, and promoting a culture of diversity and inclusion. Each ERC must partner with an independent evaluation team to annually assess their performance and impact on progressing education, connecting community, and building diversified culture. This evaluation is currently performed independently (and in isolation), which leads to inconsistent evaluations and a redundant investment of ERCs’ resources into such tasks (e.g. developing evaluation instruments). These isolated efforts by ERCs to quantitatively evaluate their education programs also typically lack adequate sample size within a single center, which limits the validity and reliability of the quantitative analyses. Three ERCs, all associated with a large southwest university in the United States, worked collaboratively to overcome sample size and measure inconsistency concerns by developing a common quantitative instrument that is capable of evaluating any ERC’s education and diversity impacts. The instrument is the result of a systematic process with comparing and contrasting each ERC’s existing evaluation tools, including surveys and interview protocols. This new, streamlined tool captures participants’ overall experience as part of the ERC by measuring various constructs including skillset development, perception of diversity and inclusion, future plans after participating in the ERC, and mentorship received from the ERC. Scales and embedded items were designed broadly for possible use with both yearlong (e.g. graduate and undergraduate student, and postdoctoral scholars) and summer program (Research Experience for Undergraduates, Research Experience for Teachers, and Young Scholar Program) participants. The instrument was distributed and tested during Summer 2019 with participants in the summer programs from all three ERCs. The forthcoming paper will present the new common cross-ERC evaluation instrument, demonstrate the effort of collecting data across all three ERCs, present preliminary findings, and discuss collaborative processes and challenges. The preliminary implication for this work is the ability to directly compare educational programs across ERCs. The authors also believe that this tool can provide a fast start for new ERCs on how to evaluate their educational programs. 
    more » « less
  3. This research paper reports the in-progress validation of a quantitative instrument designed to assess the perceived impact of participating in a National Science Foundation (NSF)-funded Engineering Research Center (ERC). A multi-institutional consortium composed of ERC education directors, researchers, and evaluators from six NSF-funded ERCs designed easily accessible evaluation instruments and tools that specifically help measure anticipated outcomes for ERC participants for all ERCs. The total effort underway by the consortium includes creating a suite of qualitative and quantitative instruments, an evaluator toolkit, and a user-friendly online platform to host the inventory materials. This paper focuses on the quantitative instrument created to evaluate the experiences of those who engage with a center. It consists of Likert-type questions assessing the impact of the ERC on participants' self-reported: 1) understanding of the ERC, 2) research and communication skills, 3) climate of inclusion, 4) mentorship experiences, and 5) program satisfaction. The instrument also included additional demographic questions and questions to capture STEM-related future plans. The instrument was designed using multiple rounds of design iterations and pilot tests. Separate surveys used by individual ERCs were compiled and categorized to ensure all requirements from the National Science Foundation were met. The web-based survey was administered to six ERCs during the Summer of 2021, Fall of 2021, and Spring of 2022. A total of 549 responses were collected; 535 were used following data cleaning procedures. Sample sizes for each component of the survey varied because some ERCs chose to only use some parts of the new instrument. Exploratory Factor Analyses (EFA) were performed to identify latent factors and items that needed further revision. The following factors emerged from our analyses: 1) ERC general understanding; 2) development of research skills; 3) development of professional skills; 4) experience in the ERC; 5) feelings toward the ERC; 6) Beliefs about the ERC, 7) mentors performance; and 8) mentorship experience. The results provide preliminary evidence that the survey can be used across ERCs. This effort is the first that has been undertaken to develop a shared ERC instrument. The data collected was used to continue in-progress validation. The collaborative nature of this effort can provide ways for ERCs to benchmark impacts of their efforts and share effective practices across ERCs and other similarly structured STEM centers going forward. 
    more » « less
  4. This Innovative Practice Work in Progress paper presents the collaborative efforts made by three NSF-funded Engineering Research Centers (ERCs) to synthesize common tools for educational program evaluation. The aim of the NSF ERCs is to achieve transformative changes by integrating engineering research and education with technological innovation within areas at the frontiers of science and engineering (e.g., NSF's 10 Big Ideas). Such centers across the nation study and innovate within their technical area using similar structures and implementation strategies, including the coordination of educational endeavors. Independent partners are enlisted as part of these centers to evaluate education and diversity impacts annually. Each center typically performs this task in isolation from other such centers. The effort required to create resources for such evaluation outcome can result in redundancy and an inability for psychometric analysis due to small available populations within a single center. This paper elaborates on the ongoing efforts of this collaborative research aimed at addressing these issues by creating a streamlined, customizable, and standardized set of evaluation instruments that can be applied to any ERC evaluation. 
    more » « less
  5. Obeid, I. (Ed.)
    The Neural Engineering Data Consortium (NEDC) is developing the Temple University Digital Pathology Corpus (TUDP), an open source database of high-resolution images from scanned pathology samples [1], as part of its National Science Foundation-funded Major Research Instrumentation grant titled “MRI: High Performance Digital Pathology Using Big Data and Machine Learning” [2]. The long-term goal of this project is to release one million images. We have currently scanned over 100,000 images and are in the process of annotating breast tissue data for our first official corpus release, v1.0.0. This release contains 3,505 annotated images of breast tissue including 74 patients with cancerous diagnoses (out of a total of 296 patients). In this poster, we will present an analysis of this corpus and discuss the challenges we have faced in efficiently producing high quality annotations of breast tissue. It is well known that state of the art algorithms in machine learning require vast amounts of data. Fields such as speech recognition [3], image recognition [4] and text processing [5] are able to deliver impressive performance with complex deep learning models because they have developed large corpora to support training of extremely high-dimensional models (e.g., billions of parameters). Other fields that do not have access to such data resources must rely on techniques in which existing models can be adapted to new datasets [6]. A preliminary version of this breast corpus release was tested in a pilot study using a baseline machine learning system, ResNet18 [7], that leverages several open-source Python tools. The pilot corpus was divided into three sets: train, development, and evaluation. Portions of these slides were manually annotated [1] using the nine labels in Table 1 [8] to identify five to ten examples of pathological features on each slide. Not every pathological feature is annotated, meaning excluded areas can include focuses particular to these labels that are not used for training. A summary of the number of patches within each label is given in Table 2. To maintain a balanced training set, 1,000 patches of each label were used to train the machine learning model. Throughout all sets, only annotated patches were involved in model development. The performance of this model in identifying all the patches in the evaluation set can be seen in the confusion matrix of classification accuracy in Table 3. The highest performing labels were background, 97% correct identification, and artifact, 76% correct identification. A correlation exists between labels with more than 6,000 development patches and accurate performance on the evaluation set. Additionally, these results indicated a need to further refine the annotation of invasive ductal carcinoma (“indc”), inflammation (“infl”), nonneoplastic features (“nneo”), normal (“norm”) and suspicious (“susp”). This pilot experiment motivated changes to the corpus that will be discussed in detail in this poster presentation. To increase the accuracy of the machine learning model, we modified how we addressed underperforming labels. One common source of error arose with how non-background labels were converted into patches. Large areas of background within other labels were isolated within a patch resulting in connective tissue misrepresenting a non-background label. In response, the annotation overlay margins were revised to exclude benign connective tissue in non-background labels. Corresponding patient reports and supporting immunohistochemical stains further guided annotation reviews. The microscopic diagnoses given by the primary pathologist in these reports detail the pathological findings within each tissue site, but not within each specific slide. The microscopic diagnoses informed revisions specifically targeting annotated regions classified as cancerous, ensuring that the labels “indc” and “dcis” were used only in situations where a micropathologist diagnosed it as such. Further differentiation of cancerous and precancerous labels, as well as the location of their focus on a slide, could be accomplished with supplemental immunohistochemically (IHC) stained slides. When distinguishing whether a focus is a nonneoplastic feature versus a cancerous growth, pathologists employ antigen targeting stains to the tissue in question to confirm the diagnosis. For example, a nonneoplastic feature of usual ductal hyperplasia will display diffuse staining for cytokeratin 5 (CK5) and no diffuse staining for estrogen receptor (ER), while a cancerous growth of ductal carcinoma in situ will have negative or focally positive staining for CK5 and diffuse staining for ER [9]. Many tissue samples contain cancerous and non-cancerous features with morphological overlaps that cause variability between annotators. The informative fields IHC slides provide could play an integral role in machine model pathology diagnostics. Following the revisions made on all the annotations, a second experiment was run using ResNet18. Compared to the pilot study, an increase of model prediction accuracy was seen for the labels indc, infl, nneo, norm, and null. This increase is correlated with an increase in annotated area and annotation accuracy. Model performance in identifying the suspicious label decreased by 25% due to the decrease of 57% in the total annotated area described by this label. A summary of the model performance is given in Table 4, which shows the new prediction accuracy and the absolute change in error rate compared to Table 3. The breast tissue subset we are developing includes 3,505 annotated breast pathology slides from 296 patients. The average size of a scanned SVS file is 363 MB. The annotations are stored in an XML format. A CSV version of the annotation file is also available which provides a flat, or simple, annotation that is easy for machine learning researchers to access and interface to their systems. Each patient is identified by an anonymized medical reference number. Within each patient’s directory, one or more sessions are identified, also anonymized to the first of the month in which the sample was taken. These sessions are broken into groupings of tissue taken on that date (in this case, breast tissue). A deidentified patient report stored as a flat text file is also available. Within these slides there are a total of 16,971 total annotated regions with an average of 4.84 annotations per slide. Among those annotations, 8,035 are non-cancerous (normal, background, null, and artifact,) 6,222 are carcinogenic signs (inflammation, nonneoplastic and suspicious,) and 2,714 are cancerous labels (ductal carcinoma in situ and invasive ductal carcinoma in situ.) The individual patients are split up into three sets: train, development, and evaluation. Of the 74 cancerous patients, 20 were allotted for both the development and evaluation sets, while the remain 34 were allotted for train. The remaining 222 patients were split up to preserve the overall distribution of labels within the corpus. This was done in hope of creating control sets for comparable studies. Overall, the development and evaluation sets each have 80 patients, while the training set has 136 patients. In a related component of this project, slides from the Fox Chase Cancer Center (FCCC) Biosample Repository ( -facility) are being digitized in addition to slides provided by Temple University Hospital. This data includes 18 different types of tissue including approximately 38.5% urinary tissue and 16.5% gynecological tissue. These slides and the metadata provided with them are already anonymized and include diagnoses in a spreadsheet with sample and patient ID. We plan to release over 13,000 unannotated slides from the FCCC Corpus simultaneously with v1.0.0 of TUDP. Details of this release will also be discussed in this poster. Few digitally annotated databases of pathology samples like TUDP exist due to the extensive data collection and processing required. The breast corpus subset should be released by November 2021. By December 2021 we should also release the unannotated FCCC data. We are currently annotating urinary tract data as well. We expect to release about 5,600 processed TUH slides in this subset. We have an additional 53,000 unprocessed TUH slides digitized. Corpora of this size will stimulate the development of a new generation of deep learning technology. In clinical settings where resources are limited, an assistive diagnoses model could support pathologists’ workload and even help prioritize suspected cancerous cases. ACKNOWLEDGMENTS This material is supported by the National Science Foundation under grants nos. CNS-1726188 and 1925494. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. REFERENCES [1] N. Shawki et al., “The Temple University Digital Pathology Corpus,” in Signal Processing in Medicine and Biology: Emerging Trends in Research and Applications, 1st ed., I. Obeid, I. Selesnick, and J. Picone, Eds. New York City, New York, USA: Springer, 2020, pp. 67 104. [2] J. Picone, T. Farkas, I. Obeid, and Y. Persidsky, “MRI: High Performance Digital Pathology Using Big Data and Machine Learning.” Major Research Instrumentation (MRI), Division of Computer and Network Systems, Award No. 1726188, January 1, 2018 – December 31, 2021. https://www. [3] A. Gulati et al., “Conformer: Convolution-augmented Transformer for Speech Recognition,” in Proceedings of the Annual Conference of the International Speech Communication Association (INTERSPEECH), 2020, pp. 5036-5040. [4] C.-J. Wu et al., “Machine Learning at Facebook: Understanding Inference at the Edge,” in Proceedings of the IEEE International Symposium on High Performance Computer Architecture (HPCA), 2019, pp. 331–344. [5] I. Caswell and B. Liang, “Recent Advances in Google Translate,” Google AI Blog: The latest from Google Research, 2020. [Online]. Available: [Accessed: 01-Aug-2021]. [6] V. Khalkhali, N. Shawki, V. Shah, M. Golmohammadi, I. Obeid, and J. Picone, “Low Latency Real-Time Seizure Detection Using Transfer Deep Learning,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2021, pp. 1 7. https://www.isip. [7] J. Picone, T. Farkas, I. Obeid, and Y. Persidsky, “MRI: High Performance Digital Pathology Using Big Data and Machine Learning,” Philadelphia, Pennsylvania, USA, 2020. [8] I. Hunt, S. Husain, J. Simons, I. Obeid, and J. Picone, “Recent Advances in the Temple University Digital Pathology Corpus,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2019, pp. 1–4. [9] A. P. Martinez, C. Cohen, K. Z. Hanley, and X. (Bill) Li, “Estrogen Receptor and Cytokeratin 5 Are Reliable Markers to Separate Usual Ductal Hyperplasia From Atypical Ductal Hyperplasia and Low-Grade Ductal Carcinoma In Situ,” Arch. Pathol. Lab. Med., vol. 140, no. 7, pp. 686–689, Apr. 2016. 
    more » « less