examples/dynamic_decisions/delayed_impact.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Delayed Impact \n",
    "\n",
    "This notebook explores causal inference in the context of decision making with dynamics. As our starting point, we use the lending simulator introduced in [Delayed Impact of Fair Machine Learning](https://arxiv.org/abs/1803.04383). \n",
    "\n",
    "To connect this simulator with causal inference, we exploit it's representation as a structural causal model introduced in [Causal Modeling for Fairness in Dynamical Systems](https://arxiv.org/abs/1909.09141). All of the interventions we consider, including the \"credit bureau\" intervention, were introduced by Creager et al. For more details, see\n",
    "\n",
    "1. Creager, Elliot, et al. \"Causal Modeling for Fairness in Dynamical Systems.\" International Conference on Machine Learning. 2020.\n",
    "\n",
    "2. Liu, Lydia, et al. \"Delayed Impact of Fair Machine Learning.\" International Conference on Machine Learning. 2018."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "import matplotlib.pylab as plt\n",
    "from tqdm.auto import tqdm\n",
    "\n",
    "import whynot as wn\n",
    "import whynot.traceable_numpy as np\n",
    "\n",
    "from whynot.simulators.delayed_impact.simulator import INV_CDFS, GROUP_SIZE_RATIO"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Constructing the experiment"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For illustrative purposes, we show how to construct the credit bureau intervention from [Creager et al.](https://arxiv.org/abs/1909.09141) in WhyNot. The user writes a handful of small functions, each determining a different aspect of the causal experiment."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Sample initial states\n",
    "Each state in the delayed impact model consists of an agent parameterized by a binary \"group\" indicator and a credit score.\n",
    "We sample group membership and credit scores based on historical FICO data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sample_initial_states(rng):\n",
    "    group = int(rng.uniform() < GROUP_SIZE_RATIO[1])\n",
    "    # Compute credit score via inverse CDF trick\n",
    "    score = INV_CDFS[group](rng.uniform())\n",
    "    return wn.delayed_impact.State(group=group, credit_score=score)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Set up simulator configuration\n",
    "The simulator is parameterized by lending thresholds for each group $\\tau_0$ and $\\tau_1$, which we instantiate as `parameters` to allow the user to vary them\n",
    "during experiments."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "@wn.parameter(\n",
    "    name=\"threshold_g0\", default=550, description=\"Lending threshold for group 0\")\n",
    "@wn.parameter(\n",
    "    name=\"threshold_g1\", default=550, description=\"Lending threshold for group 1\")\n",
    "def construct_config(threshold_g0, threshold_g1):\n",
    "    \"\"\"Return the experimental config for runs without intervention\"\"\"\n",
    "    return wn.delayed_impact.Config(\n",
    "        start_time=0, end_time=1, threshold_g0=threshold_g0, threshold_g1=threshold_g1\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Define outcome measurement\n",
    "The outcome measurement we measure is the change in an agent's credit score from time step 0 to time step 1, along with the profit\n",
    "the bank earns for this individual."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def extract_outcomes(run):\n",
    "    # Recall states are individuals in this model\n",
    "    agent_t0 = run.states[0]\n",
    "    agent_t1 = run.states[1]\n",
    "    return [agent_t1.credit_score - agent_t0.credit_score, agent_t1.profits]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Define the intervention\n",
    "\n",
    "The intervention we consider is the implementation of a credit bureau that treats intermediates between the individual and the lender\n",
    "and reports credit scores of $\\min(\\text{score}, 600)$, as discussed in [Creager et al.](https://arxiv.org/abs/1909.09141)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def creditscore_threshold(score):\n",
    "    \"\"\"Alternate credit bureau scoring policy.\"\"\"\n",
    "    return max(score, 600)\n",
    "\n",
    "def intervention():\n",
    "    return wn.delayed_impact.Intervention(credit_scorer=creditscore_threshold, time=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Put the components together\n",
    "We put each of these components in a `DynamicsExperiment` object to create an the `CreditBureauExperiment`. Importantly, we assume all agents in the model\n",
    "have the credit bureau intervention, so they are \"treated\" in the causal inference sense with probability 1.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "CreditBureauExperiment = wn.DynamicsExperiment(\n",
    "    name=\"CreditBureauExperiment\",\n",
    "    description=\"Intervention on the credit scoring mechanism.\",\n",
    "    simulator=wn.delayed_impact,\n",
    "    simulator_config=construct_config,\n",
    "    intervention=intervention,\n",
    "    state_sampler=sample_initial_states,\n",
    "    propensity_scorer=1.0,\n",
    "    outcome_extractor=extract_outcomes,\n",
    "    covariate_builder=lambda run: run.initial_state.group,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Running the experiment\n",
    "\n",
    "Compute average credit score changes for the minority group, as well as institutional profits, as the credit score threshold for the minority varies from 300 to 800.\n",
    "Throughout, the majority threshold is fixed."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b145acd6457a4bfb8afa98691e8d5fb3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=50.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "minority_thresholds = list(range(300, 800, 10))\n",
    "\n",
    "average_min_score_changes = []\n",
    "average_inst_profits = []\n",
    "for tau_0 in tqdm(minority_thresholds):\n",
    "    # Run the experiment to generate the dataset\n",
    "    dataset = CreditBureauExperiment.run(threshold_g0=tau_0, num_samples=1000, parallelize=True)    \n",
    "    \n",
    "    # Only consider score changes for the minority group\n",
    "    minority_locs = dataset.covariates[:, 0] == 0\n",
    "    minority_treated_locs = minority_locs & (dataset.treatments == 1)\n",
    "\n",
    "    score_changes = dataset.outcomes[:, 0]\n",
    "    minority_changes = score_changes[minority_treated_locs]\n",
    "    average_min_score_changes.append(np.mean(minority_changes))\n",
    "        \n",
    "    # Report profits over the entire group\n",
    "    ind_profits = dataset.outcomes[:, 1][dataset.treatments == 1]\n",
    "    average_inst_profits.append(np.mean(ind_profits))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualizing the results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "_, axs = plt.subplots(1, 2, figsize=(12, 6))\n",
    "axs[0].plot(minority_thresholds, average_min_score_changes, label=\"Average Minority\\n Score Change\")\n",
    "axs[0].plot([600] * 20, np.linspace(-60, 20, 20), label=\"Credit Bureau Cutoff\", linestyle=\"--\", color=\"black\")\n",
    "axs[0].legend()\n",
    "axs[0].set_xlabel(\"Minority Threshold\")\n",
    "axs[0].set_ylabel(\"Score Change\")\n",
    "axs[1].plot(minority_thresholds, average_inst_profits, label=\"Average Profit\")\n",
    "axs[1].plot([600] * 20, np.linspace(0, 0.5, 20), label=\"Credit Bureau Cutoff\", linestyle=\"--\", color=\"black\")\n",
    "axs[1].set_xlabel(\"Minority Threshold\")\n",
    "axs[1].set_ylabel(\"Institutional Profit\")\n",
    "axs[1].legend();"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Treatment effect estimation with confounding\n",
    "\n",
    "Now, we slightly modify the CreditBureau example to show how WhyNot can also be used to address causal inference questions. Rather than exposing all agents to the credit bureau intervention, we imagine\n",
    "members of the minority group are more likely to receive the credit bureau intervention (75% chance) compared to members of the majority group (20% chance). Using this data, we wish to estimate the population-level causal effect of the credit intervention.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Constructing the experiment\n",
    "\n",
    "We slightly modify the `CreditBureauExperiment` to include confounded treatment assignment by changing the ``propensity_scorer`` function."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def propensity_scorer(untreated_run):\n",
    "    \"\"\"Assign minority to treatment 75% of the time, compared to 25% for the majority\"\"\"\n",
    "    return  0.75 * (untreated_run.initial_state.group == 0) + 0.2 * (untreated_run.initial_state.group == 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "BiasedCreditBureauExperiment = wn.DynamicsExperiment(\n",
    "    name=\"BiasedCreditBureauExperiment\",\n",
    "    description=\"Intervention on the credit scoring mechanism, with treatment bias.\",\n",
    "    simulator=wn.delayed_impact,\n",
    "    simulator_config=construct_config,\n",
    "    intervention=intervention,\n",
    "    state_sampler=sample_initial_states,\n",
    "    propensity_scorer=propensity_scorer,\n",
    "    outcome_extractor=extract_outcomes,\n",
    "    # Only covariate is group membership, which is a confounder for this experiment.\n",
    "    covariate_builder=lambda run: [run.initial_state.group, run.initial_state.credit_score]\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Running the experiment and generating causal graphs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Run the experiment to generate an observational dataset. Using `causal_graph=True` generates the causal graph associated with the experiment."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1d45078b084c48b692f55d29076bf63c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=1000.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "dataset = BiasedCreditBureauExperiment.run(num_samples=1000, causal_graph=True, show_progress=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Estimating treatment effects"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Estimated ATE: -26.07 (-35.41, -16.74)\n",
      "True ATE: -24.74\n"
     ]
    }
   ],
   "source": [
    "covariates, treatment, outcome = dataset.covariates, dataset.treatments, dataset.outcomes\n",
    "score_changes = dataset.outcomes[:, 0]\n",
    "\n",
    "inference_result = wn.algorithms.ols.estimate_treatment_effect(\n",
    "    covariates, treatment, score_changes)\n",
    "\n",
    "print(\"Estimated ATE: {:.2f} ({:.2f}, {:.2f})\".format(inference_result.ate, *inference_result.ci))\n",
    "print(\"True ATE: {:.2f}\".format(np.mean(dataset.true_effects[:, 0])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Inspecting the causal graph\n",
    "\n",
    "The causal graph is a `networkx.Digraph` that can easily be connected to graphical methods for estimating treatment effects, e.g. [DoWhy](https://github.com/microsoft/dowhy). \n",
    "For a complete example, see [here](https://github.com/zykls/whynot/blob/master/examples/causal_inference/graphical_methods.ipynb)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "## NODES ##\n",
      "group_0\n",
      "credit_score_0\n",
      "profits_0\n",
      "loan_approved_0\n",
      "repaid_0\n",
      "PARAM:credit_scorer_0\n",
      "PARAM:threshold_g0_0\n",
      "PARAM:threshold_g1_0\n",
      "PARAM:repayment_utility_0\n",
      "PARAM:default_utility_0\n",
      "PARAM:repayment_score_change_0\n",
      "PARAM:default_score_change_0\n",
      "PARAM:min_score_0\n",
      "PARAM:max_score_0\n",
      "group_1\n",
      "credit_score_1\n",
      "profits_1\n",
      "loan_approved_1\n",
      "repaid_1\n",
      "PARAM:credit_scorer_1\n",
      "PARAM:threshold_g0_1\n",
      "PARAM:threshold_g1_1\n",
      "PARAM:repayment_utility_1\n",
      "PARAM:default_utility_1\n",
      "PARAM:repayment_score_change_1\n",
      "PARAM:default_score_change_1\n",
      "PARAM:min_score_1\n",
      "PARAM:max_score_1\n",
      "Treatment\n",
      "Outcome\n",
      "\n",
      "## EDGES ##\n",
      "\n",
      "('group_0', 'group_1')\n",
      "('group_0', 'credit_score_1')\n",
      "('group_0', 'profits_1')\n",
      "('group_0', 'loan_approved_1')\n",
      "('group_0', 'repaid_1')\n",
      "('group_0', 'Treatment')\n",
      "('credit_score_0', 'credit_score_1')\n",
      "('credit_score_0', 'profits_1')\n",
      "('credit_score_0', 'loan_approved_1')\n",
      "('credit_score_0', 'Outcome')\n",
      "('profits_0', 'profits_1')\n",
      "('PARAM:credit_scorer_0', 'credit_score_1')\n",
      "('PARAM:credit_scorer_0', 'profits_1')\n",
      "('PARAM:credit_scorer_0', 'loan_approved_1')\n",
      "('PARAM:threshold_g0_0', 'credit_score_1')\n",
      "('PARAM:threshold_g0_0', 'profits_1')\n",
      "('PARAM:threshold_g0_0', 'loan_approved_1')\n",
      "('PARAM:threshold_g1_0', 'credit_score_1')\n",
      "('PARAM:threshold_g1_0', 'profits_1')\n",
      "('PARAM:threshold_g1_0', 'loan_approved_1')\n",
      "('PARAM:repayment_utility_0', 'profits_1')\n",
      "('PARAM:default_utility_0', 'profits_1')\n",
      "('PARAM:repayment_score_change_0', 'credit_score_1')\n",
      "('PARAM:default_score_change_0', 'credit_score_1')\n",
      "('PARAM:min_score_0', 'credit_score_1')\n",
      "('PARAM:max_score_0', 'credit_score_1')\n",
      "('credit_score_1', 'Outcome')\n",
      "('Treatment', 'PARAM:credit_scorer_0')\n",
      "('Treatment', 'PARAM:credit_scorer_1')\n"
     ]
    }
   ],
   "source": [
    "graph = dataset.causal_graph\n",
    "print(\"## NODES ##\")\n",
    "for node in graph.nodes:\n",
    "    print(node)\n",
    "print(\"\\n## EDGES ##\\n\")\n",
    "for edge in graph.edges:\n",
    "    print(edge)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}