User Modeling - Science topic

I am exploring the concept of dynamically modeling user identity and intent by analyzing behavior across generative AI tools (e.g., ChatGPT, Claude, Notion AI, etc.). The goal is to infer high-level cognitive or professional trajectories based on a user’s queries and interaction patterns, without directly accessing the raw prompt content (for example, utilizing embeddings, token patterns, and usage metadata).

I’m particularly curious about the following key areas:

1. Has there been research on creating intent graphs or behavioral fingerprints using data from interactions with large language models (LLMs)?

2. Are there known architectures or agent-based frameworks for building persistent identity models that work across multiple tools and platforms?

3. What approaches are available (or emerging) regarding zero-knowledge sharing or federated learning to ensure that user modeling remains private?

I am especially interested in applying this work to professional development or networking. However, I welcome insights from anyone researching multi-agent learning, identity in AI systems, or private user modeling. I am open to papers, frameworks, or even counterarguments related to these topics.

Hello everyone,

I want to use a Fortran Subroutine for biomass gasification to implement improved kinetics for tar formation (kinetics from Abdelouahed.2012) using a RPLUG reactor. While the subroutines for conventional components work well, I am having trouble accessing and changing properties of the biomass (as a non-conventional component) in the subroutine. I tried using the stream structure from the "Aspen Plus User Models"-help, which should result in one array for all substreams (conventional, CISOLID, and non-conventional) listing each component flows and physical properties of each substream one after another. But after all data of the conventional substream, there are no more values in the array where the data of CISOLD and non-conventional substreams should be. So, how can I access the values? Do I need to include an additional command? It also does not seem to be a matrix structure (column for each substream).

Thanks a lot for your help!

I have three modules (in free form .f90 format) which are being called from inside of a UMAT subroutine in ABAQUS, in the following manner:

module module_A

use module_C

use module_B

....

end module_A

module module_B

use module_C

....

end module_B

module module_C

....

end module_C

subroutine UMAT(STRESS,...)

....

Here the subroutines from module_A and module_B are being called

...

end subroutine UMAT

Now, what is the appropriate format for writing these modules with UMAT subroutine? How to merge different module files into a single *.for file (free format)?

Our recent research shows that AR systems have inherent conflict while interacting with virtual objects. We termed this new conflict as Virtual Kinesthetic Conflict (VKC). This conflict is very similar to the inherent Visual Accommodation Conflict (VAC) in VR. Just like VAC, VKC also cannot be avoided, we can only reduce the effects of VKC. In our recent publication, we have listed a few guidelines to reduce the effects of VKC. Can you think of other solutions?

Hi all,

I have been wracking my brain and searching the literature for answers, but yet none seem clear to me so I feel I must ask. Are "fit statistics (e.g., RMSEA, TLI, CFI, etc..) important or even useful for mediation models?

So my issue is as follows:

When I specify the following model in Lavaan, using DWLS estimation (due to ordinal factors and mediators):

outcome ~ a*fact1+fact2+fact3+fact4+fact5+fact6+b*med1+c*med2 # DIRECT EFFECT

med1 ~ a2*fact1+fact2+fact3+fact4+fact5+fact6

med2 ~ a3*fact1+fact2+fact3+fact4+fact5+fact6

#indirect effect(s)

indirect_mod1:=a2*b

indirect_mod2:=a3*c

#total effect(s)

total_crp:=a +(b*c)

I fail to meet the thresholds for the fit statistics, but the modification indicies suggest allowing covariance between mod1 and mod2, or one predicting the other (which technically could make sense), however, this then renders my model saturated and the fit statistics become useless.

Any ideas what I can do please? The output of the model is interesting, but I am concerned that the fit is too far off, but can't think of any other way to assess mediators without SEM. I have tried looking at the model-implied (fitted) covariance matrix and the residuals of a fitted model, but I cant see anything drastically out of the ordinary there.

Below are Fit outputs for both the model before mediator covariance and after, for reference.

lavaan 0.6-7 ended normally after 53 iterations

Estimator DWLS

Optimization method NLMINB

Number of free parameters 54

Number of observations 9619

Model Test User Model:

Test statistic 35.907

Degrees of freedom 1

P-value (Chi-square) 0.000

Model Test Baseline Model:

Test statistic 474.621

Degrees of freedom 24

P-value 0.000

User Model versus Baseline Model:

Comparative Fit Index (CFI) 0.923

Tucker-Lewis Index (TLI) -0.859

Root Mean Square Error of Approximation:

RMSEA 0.060

90 Percent confidence interval - lower 0.044

90 Percent confidence interval - upper 0.078

P-value RMSEA <= 0.05 0.138

Standardized Root Mean Square Residual:

SRMR 0.009

Parameter Estimates:

Standard errors Bootstrap

Number of requested bootstrap draws 1000

Number of successful bootstrap draws 1000

lavaan 0.6-7 ended normally after 54 iterations

Estimator DWLS

Optimization method NLMINB

Number of free parameters 55

Number of observations 9619

Model Test User Model:

Test statistic 0.000

Degrees of freedom 0

Model Test Baseline Model:

Test statistic 474.621

Degrees of freedom 24

P-value 0.000

User Model versus Baseline Model:

Comparative Fit Index (CFI) 1.000

Tucker-Lewis Index (TLI) 1.000

Root Mean Square Error of Approximation:

RMSEA 0.000

90 Percent confidence interval - lower 0.000

90 Percent confidence interval - upper 0.000

P-value RMSEA <= 0.05 NA

Standardized Root Mean Square Residual:

SRMR 0.000

Parameter Estimates:

Standard errors Bootstrap

Number of requested bootstrap draws 1000

Number of successful bootstrap draws 1000

Thank you so much for taking the time to read my question

Currently, I am running a model with 6 variables among which 2 are dependent. I also have from three to five items per variable (overall, 23 items). The results of the model fit for measurement model are pretty good except for the chi-square value (302.149), df=213, p-value = .000. Therefore, since p-value appears to be insignificant, I am wondering what procedures should I follow in this case? Also, whether eliminating the excessive number of items can help to solve this? If so, what parameters apart from factor loadings do I need to check?

Bifactor models allow users to model both a general factor and several specific (or group) factors. The focus is usually on the general factor, which is what the measure was designed for in the first place. However, we usually don't model CFAs every time we test a hypothesis; instead, we tend to rely on bifactor CFAs to support the creation of composite measures (i.e. the average or sum of all the items).

My question is: What's the best reliability estimate of such composite measures? Or, alternatively, what should I take as measurement error in a bifactor model? All the variance not explained by all the factors together (i.e. omega total reliability) or only the variance that cannot be attributed to the general factor (i.e. omega hierarchical reliability)?

As an example of the practical implications of this issue, consider the following side question: What reliability estimate (omega total or omega hierarchical) should we use when performing correction for attenuation?

HI everyone,

I am trying to find Ontologies for user modelling and for disabilities, so far I have had little success finding ontology search engines. The only one I managed to locate is https://www.ebi.ac.uk/ols/index which doesn't have the kinds of Ontologies I am looking for. Any pointers will be appreciated.

Thank you.

Nigeria' s subsistence farmers who constitute over 95% of farmers in the country contributing about 90% of farming produce are resource poor. They make hard decisions prior to use of new technologies of farm production knowing that failure of such practices will come with dire consequences. It is therefore expected that the farmers will want to use solutions to solve their basic farm problems that do not require much effort in terms of cost, energy and time. It is my belief that Lazy user theory could be used to provide the blueprint necessary impetus to boost agricultural production in the country

I am looking for the user specific data/log files from the serious games to use it for creating learner profiles. 

I am trying to understand if there are existing Recommender Systems which Personalizes a User Model bases on Cognitive Variable and VALs (Values, Attitude, Lifestyles).  

Are there any standard Conceptual/Concrete Personalization User Model which accounts for Cognitive variables?

Is anyone working in this area?

Thanks!

Jay

Which are the boundaries (if we can define them) that we need to take into account when we talk about User Model rather than User-context? Could we define the former more comprehensively than the latter? Or could we say that one includes the other?

Any opinion, or more precise research works I can refer to, is welcome!

Few of these tools support transformations from use case models to another analysis or design model. What is your opnion on why?